Patent Grant
7362805
1. Field of the Invention
This invention relates to a digital signal recording/reproducing apparatus such as a video tape recorder (hereinafter, abbreviated as “VTR”), a video disk player and an audio tape recorder in which video and audio signals are recorded and reproduced in the digital form, and particularly to an apparatus which performs motion-compensation prediction on a video signal for compression-encoding.
2. Description of the Related Art
In a digital VTR for home use, data compression is indispensable in view of cost and hardware size. Hereinafter, therefore, data compression will be described taking mainly a digital VTR for home use as an example.
Next, the data compressor (high-efficiency encoder) 902 will be described.
Reference numeral 12 designates an inverse quantizing circuit which performs inverse-quantization on the quantized DCT coefficient, 13 designates an inverse DCT circuit which performs the inverse DCT on the inverse-quantized DCT coefficient, 14 designates an adder which adds a prediction block to the decoded block that is an output signal of the inverse DCT circuit 13 to generate an output block, 15 designates a video memory which stores output blocks in order to perform motion-compensation prediction, 16 designates an MC circuit which performs motion estimation from a motion-compensation search block segmented from a past image stored in the video memory 15 and the current input block, and performs motion-compensation prediction, 17 designates a MIX circuit which combines a motion vector with a mode signal determined by the determiner 6, 18 designates a second encoder which codes the output of the MIX circuit 17, and 19 designates a second switch which switches the prediction blocks in accordance with the mode determined by the determiner 6. The difference power calculator 4, original power calculator 5, determiner 6, inverse quantizing circuit 12, inverse DCT circuit 13, adder 14, video memory 15, MC circuit 16 and second switch 19 constitute a local decoding loop 20.
Then, the operation will be described. Irrespective of an intra-field in which motion-compensation prediction is not performed or a prediction-field (inter-field) in which motion-compensation prediction is performed, input digital video signals are divided by the blocking circuit 2 into input blocks in the unit of m [pixels]×n [lines] (where m and n are positive integers), and segmented. In order to obtain a difference block, the subtracter 3 calculates the difference in the unit of pixel between an input block and a prediction block. Then) the input block and the difference block are input into first switch 7. The difference power calculator 4 calculates the difference power of the difference block. On the other hand, the original power calculator 5 calculates the original AC power of the input block. The two calculated powers are compared with each other by the determiner 6 to control the first switch 7 so that the block having the smaller power is selected as the encoding subject. More specifically, when the difference power is smaller than the original AC power, the determiner 6 outputs a prediction mode signal, and in contrast with this, when the original AC power is smaller than the difference power, the determiner 6 outputs an intra mode signal.
The first switch 7 outputs the input block or the difference block as an encoded block in accordance with the mode signal determined by the determiner 6. When the processed image is in the intra-field, however, the first switch 7 operates so that all of the encoded blocks are output as input blocks.
The encoded block selected by the first switch 7 is converted into DCT coefficients by the DCT circuit 8, and then subjected to the weighting and threshold processes in the quantizing circuit 9 to be quantized to predetermined bit numbers respectively corresponding to the coefficients. The quantized DCT coefficients are converted by the first encoder 10 into codes suitable for the transmission path 11 and then output to the transmission path 11.
The quantized DCT coefficients also enter into the local decoding loop 20, and the image reproduction for next motion-compensation prediction is performed. The quantized DCT coefficients which have entered into the local decoding loop 20 are subjected to the inverse weighting and inverse quantizing processes in the inverse quantizing circuit 12. Then, the DCT coefficients are converted into a decoded block by inverse DCT circuit 13. The adder 14 adds the decoded block to a prediction block in the unit of pixel to reconstruct the image. This prediction block is the same as that used in the subtracter 3. The output of the adder 14 is written as an output block in a predetermined address of the video memory 15. The memory capacity of the video memory 15 depends on the type of the employed predictive method. Assuming that the video memory 15 consists of a plurality of field memories, the reconstructed output block is written in a predetermined address. A block which is segmented from an image reconstructed from past output blocks and is in the motion estimation search range is output from the video memory 15 to the MC circuit 16. The size of the block in the motion estimation search range is i [pixels]×j [lines] (where i≧m, j≧n, and i and j are positive integers). Data in the search range from the video memory 15 and an input block from the blocking circuit 2 are input to the MC circuit 16 as data, thereby extracting motion vectors. As a method of extracting motion vectors, there are various methods such as the total search block matching method, and the tree search block matching method. These methods are well known, and therefore their description is omitted.
The motion vectors extracted by the MC circuit 16 are input to the MIX circuit 17, and combined therein with the mode signal determined by the determiner 6. The combined signals are converted by the second encoder 18 into codes suitable for the transmission path 11, and then output together with the corresponding encoded block to the transmission path 11. The MC circuit 16 outputs as a prediction block signals which are segmented from the search range in the size (m [pixels]×n [lines]), which is equal to that of the input block. The prediction block to be output from the MC circuit 16 is produced from past video information. The prediction block is supplied to second switch 19, and output from the respective output terminal of the switch in accordance with the field of the currently, processed image and the mode signal of the decoded block. Namely, the prediction block is output from one of the output terminals of the second switch 19 to the subtracter 3 in accordance with the processed field, and from the other output terminal in accordance with the mode signal of the current decoded block and the processed field.
As a predictive method used in such a circuit block, for example, the method shown in
Initially, first field F1 is blocked in the field and subjected to the DCT. Then, first field F1 is subjected to the weighting and threshold processes and quantized, and thereafter encoded. In the local decoding loop 20, the quantized signals of first field F1 are decoded or reconstructed. The reconstructed image is used in motion-compensation prediction for second and third fields F2 and F3. Then, motion-compensation prediction’ is performed on second field F2 using first field F1. After the obtained difference block is subjected to the DCT, encoding is performed in a similar manner as in first field F1. In this case, when the AC power of the input block is smaller than the power of the difference block, the input block in place of the difference block is subjected to the DCT, and thereafter encoding is performed in a similar manner as in first field F1. Second field F2 is decoded and reconstructed in the local decoding loop 20 in accordance with the mode signal of each block, and then used in motion-compensation prediction for fourth field F4. In a similar manner as in second field F2, using first field F1, motion-compensation prediction and encoding are performed on third field F3. Motion-compensation prediction is performed on fourth field F4 using second field F2 reconstructed in the video memory 15, and then, fourth field F4 is encoded in a similar manner as in third field F3. Also in third and fourth fields F3 and F4, when the AC power of the input block is smaller than the power of the difference block, the input block in place of the difference block is subjected to the DCT, and thereafter encoding is performed in a similar manner as in first field F1.
For example, the digital VTR for home use shown in
In such a predictive method, since motion-compensation prediction is performed using the video data of the one preceding field or frame, there arises a first problem that the capacity of the field memory or frame memory is increased and the hardware is enlarged in size.
In the conventional predictive method, when a scene change once occurs in the unit of frame, it is difficult during encoding of the image after the scene change to perform the compression according to motion-compensation prediction from the reference picture which was obtained before the scene change, thereby causing a second problem that the total amount of codes is increased. If the inter-frame motion-compensation prediction is performed on the whole sequentially in the temporal direction, it may be possible to suppress the increase in the data amount to a minimum level even when a scene change occurs. In the case of encoding interlace images without scene change and with less motion, however, there is a tendency that the data amount is increased as a whole. In a predictive method in which third and fourth fields F3 and F4 are adaptively switched from first, second and third fields F1, F2 and F3 as shown in
In the case that the encoding is done by performing prediction as in the prior art,
As conventional VTRs for home use of helical scanning type, there are VHS type, β type and 8-mm type VTRs. Hereinafter, a VTR of 8-mm type will be described as an example of a prior art.
In
Next, a specific example of the operation of a conventional apparatus will be described with reference to
An audio signal given to an audio signal input terminal 202 is supplied through the time-base compressing circuit 209, a modulating circuit 210 and a switch 241 for switching between the recording and the erasing, to the third and fourth gate circuits 211 and 212. The output of the erasing current generator 240 is supplied through the switch 241 to the third and fourth gate circuits 211 and 212. The output signals of the third and fourth gate circuits 211 and 212 are supplied to the adders 213 and 214, respectively. The output signal of the adder 213 is given to a rotary transformer 217 through a changeover switch 215 for switching between the recording and the erasing. The output signal of the rotary transformer 217 is given to a rotary magnetic head 221 through a rotation shaft 219 and a rotary head bar 220, so that a recording current or an erasing current flows into a magnetic tape 223.
The output signal of the adder 214 is given to a rotary transformer 218 through a switch 216 which is used for switching between recording and the erasing and is interlocked with the switch 215. The output signal of the rotary transformer 218 is given to another rotary magnetic head 222 through the rotation shaft 219 and the rotary head bar 220, so that a recording current or an erasing current flows into the magnetic tape 223. The magnetic tape 223 is guided by guide posts 224 and 225 placed on the either sides of a table guide drum 226 which has rotary magnetic heads 221 and 222 built in, and is run at a constant speed in the direction of arrow 227, by a well known magnetic tape running device (not shown) which consists of capstans and pinch rollers. The table guide drum 226 may have a well-known structure, and therefore its specific description is omitted.
In the reproduction process, a signal reproduced by the rotary magnetic head 221 is supplied to a separating circuit 228 through the rotary head bar 220, the rotation shaft 219, the rotary transformer 217 and the switch 216. On the other hand, a signal reproduced by the rotary magnetic head 222 is supplied to a separating circuit 229 through the rotary head bar 220, the rotation shaft 219, the rotary transformer 218 and the switch 216. One of the outputs of the separating circuit 228 and one of the outputs of the separating circuit 229 are supplied to an adder 230. The other output of the separating circuit 228 and the other output of the separating circuit 229 are supplied to an adder 231. The output signal of the adder 230 is supplied to a video signal output terminal 233 through a video signal processing circuit 232. In contrast, the output signal of the adder 231 is supplied to an audio signal output terminal 237 through time-base correcting circuit 234, a demodulating circuit 235 and a time-base expanding circuit 236.
Then, the operation will be described. A video signal given to the video signal input terminal 201 is converted into an FM signal by the video signal processing circuit 203. When the video signal includes a chrominance signal, the chrominance signal is converted into a low frequency signal of less than about 1.2 MHz. There will be no problem that, for example, the phase of the chrominance signal is shifted by 90 deg. or inverted every 1H (horizontal scanning interval) as means for eliminating an adjacent color signal. This is a technique of eliminating a crosstalk between tracks with using the line correlation of chrominance signal. Such processed video signal is supplied to first and second gate circuits 205 and 206.
On the other hand, since the video signal is given also to the synchronizing signal separating circuit 204, a vertical synchronizing signal is obtained at the end of the output of the circuit. The vertical synchronizing signal is supplied to the delay circuits 207 and 208. The delay circuit 207 has functions of dividing an input signal into a half frequency and delaying a signal. From the ends of Q and
The erasing current generator 240 generates an erasing current of a certain frequency (for example, 100 kHz). The timing of starting the oscillation of the erasing current is controlled by a trigger signal T which is obtained by delaying the vertical synchronizing signal in the delay circuit 208. The erasing current is output through the switch 241 to the third and fourth gate circuits 211 and 212 to which the pulse signals Q and
During the reproduction process, the moving contacts of the switches 215 and 216 are positioned at fixed contacts P. This allows the two-channel reproduction signal reproduced by the rotary magnetic heads 221 and 222 to be respectively transmitted through the rotary head bar 220, the rotation shaft 219, the rotary transformers 217 or 218 and the switches 215 or 216, and to be respectively separated into a video signal and an audio signal on the time-base in separating circuits 228 and 229. The separated video signals are converted by the adder 230 into a one-channel video signal which is continuous in terms of time, and then supplied to the video signal processing circuit 232. The video signal processing circuit 232 reconstructs the original video signal from the input signal, and outputs the reconstructed signal to the video signal output terminal 233. On the other hand, the separated audio signals are converted into a one-channel of signal by the adder 231, and then supplied to the time-base correcting circuit 234. The time-base correcting circuit 234 consists of a semiconductor memory device such as a CCD (charge-coupled device) and a BED (bucket brigade device), and eliminates time-base variations (so-called jitter and skew distortion) of the tape and head system. The output signal of the time-base correcting circuit 234 is demodulated to the original compressed audio signal by the demodulating circuit 235. The demodulated signal is then converted into the original audio signal by the time-base expanding circuit 236 consisting of a semiconductor memory device such as a CCD and a BED, and output to the audio signal output terminal 237.
As described above, in an 8-mm VTR, video signals and audio signals for one field are recorded on and reproduced from one track on a tape.
In a digital recording, it is not necessary to record a video signal during the entire period. In a blanking interval, a video signal has a constant waveform irrespective of the contents of an image. Since this waveform can be synthesized after the reproduction, in both the D1 and D2 methods, the recording is performed only during the effective video period. Also a color burst signal appearing in a blanking interval of an NTSC signal can be synthesized after the reproduction. This is because the sampling phase in the D2 method is set to the I and Q axes and the phase of the color burst (lagging behind Q axis by (180+33) deg.) can be determined using a reproduced sampling clock.
When a video signal in a segment is to be recorded, it is divided into four channels in the D1 method, and into two channels in the D2 method. As a result, the number of pixels per one channel of one segment is {(720+360×2)/4}×50=360×50=18,000 in the D1 method, and (768/2)×85=384×85=32,640 in the D2 method. Channels are distributed so that they are uniformly dispersed on a screen. Accordingly, even when the characteristics of a specific channel are deteriorated, code errors caused by this deterioration are not concentrated on one portion of the screen so as to be inconspicuous. Therefore, the effect of correction on errors which have not been corrected is also enormous.
In both the D1 and D2 methods, two kinds of error-correction codes which are respectively called an outer code and an inner code are used together. In an actual process of generating outer and inner codes, an operation of rearranging the sequence of the codes is performed. This operation is called shuffling. The shuffling disperses the effect of code errors, improves the correction capability, and reduces the display deterioration caused by errors which have not been corrected. The shuffling process consists of the shuffling for one scanning line which is performed before the generation of an outer code, and the shuffling which is performed in one sector after the addition of an outer code and before the generation of an inner code. As described above, in a VTR of the D1 or D2 method, video signals and audio signals for one field are recorded in a plurality of tracks on a tape.
In order to record all information of standard television signals of currently used NTSC and PAL systems, in a VTR for home use, the carrier frequency of an FM luminance signal is raised and the bandwidth and deviation are increased so as to improve the resolution and C/N ratio. However, a VTR for home use still fails to match with a VTR for business use in S/N ratio, waveform reproducibility, etc. The down sizing of a VTR is highly expected to be achieved, and, there are demands for further improvement in performance as well as realization of VTR which is light and compact. Hence, it is difficult to attain the desired performance by only improving the present techniques. On the other hand, in the field of VTRs for business use and broadcasting use, rapid advance in digitalization of an apparatus has been made to achieve multifunction and high performance in the apparatuses, and most of VTRs for broadcasting use are replaced with digital VTRs. However, a digital VTR consumes a large amount of tape, which obstacles to achieve prolongation of the recording time and the down sizing.
Recently, in view of the redundancy of information contained in an image, studies on compressing recorded information have been actively conducted, and the application of the results of these studies to a VTR is being examined. It is expected to realize a VTR which is compact and light, has high image quality and can operate long-time by achieving high image quality, and reduction of tape consumption due to high density recording and information compression which are inherent in the digital recording.
The redundancy of an input video signal always varies, and therefore, the amount of codes which are compression-coded using this redundancy also varies. However, the amount of information which can be transmitted through the transmission path 114 is limited. In order to exhibit the best of the performance, the variation of the code amount is buffered using the buffer memory 113, and the information amount is controlled to be within a predetermined range so that overflow or underflow of a memory does not occur.
Since the communication apparatus is controlled as described above, the relationship between fields of an input video signal and transmitted codes is not clearly defined. Unlike an application in the field of communication, a VTR is required to perform functions peculiar to a VTR and including a special reproduction different from normal reproduction such as a still reproduction, slow reproduction and high-speed reproduction, an assemble edition, and an insert edition. Therefore, it is desirable to clearly define the relationship between fields and tracks. In order to produce a practical VTR, it is essential to select a recording format which can solve these problems.
As a method of compressing a moving picture such as a television signal, there is a method using an intra-field (or intra-frame) in which the encoding is completed within an individual field (or frame) independently of another field (or frame), and prediction-field (or prediction frame) in which the predictive encoding is performed using information of another field (or frame). Generally, the information which the prediction between fields (or frames) is not used is two or more times the code amount of the prediction-field using the prediction between planes. When record areas of the same size (number of tracks) are allocated to the intra-field (or intra-frame) and the prediction-field (or prediction frame), there arises a fourth problem that in the intra-field (or intra-frame), the record area is not sufficient and in the prediction-field (or prediction frame) the record area has a useless portion.
It is an object of the invention to provide a high-efficiency encoder which does not perform motion-compensation on an intra-field (or intra-frame) in every n fields (or n frames) and which performs motion-compensation prediction with an intra-field (or intra-frame) in other fields (or frames), thereby solving the first problem and reducing the hardware size.
It is another object of the invention to provide a high-efficiency encoder in which a reference picture is previously set using a method fixed somewhat for a normal picture having no scene change, and, when a scene change occurs in the step of the motion-compensation prediction process, the reference picture is switched so as to suppress the amount of generated information as much as possible, thereby solving the above mentioned second problem. Further, the S/N ratio of a picture can be maintained while suppressing the increase in the hardware size.
It is a further object of the invention to provide a high-efficiency encoder in which motion-compensation prediction is performed while a reference picture is previously set using a method fixed somewhat for a normal picture having no scene change, and, when a scene change occurs in the step of the motion-compensation prediction process, the picture immediately after the scene change is encoded in a field (or frame) as an intra-field (or intra-frame). Thus, the third problem mentioned above is solved and the deterioration of a picture quality is suppressed after the scene change as much as possible.
It is a still further object of the invention to provide a video information recording/reproducing apparatus in which signals of a plurality of fields (or frames) are collected into one recording unit to be recorded in a predetermined number of tracks, thereby solving the above mentioned fourth problem and coping with the special reproduction and edition required in a VTR.
In a high-efficiency encoder of the invention, only an intra-field (or intra-frame) is used as a reference picture for motion-compensation prediction, and therefore the information compression in which the deterioration of the image quality is not conspicuous can be performed with a reasonable hardware size.
In the other high-efficiency encoder of the invention, motion-compensation prediction is usually performed using a fixed reference picture, and, when a scene change occurs in the step of the motion-compensation prediction process, the reference picture is adaptively switched, thereby suppressing the increase in the information amount. In this case, for example, blocks which are judged to be in an intra mode are counted in order, to determine whether a scene change occurred, and, when blocks of an intra mode, the number of which is greater than a preset threshold, are generated, the reference picture for the next field (or frame) is switched. Therefore, the amount of generated codes can be suppressed and the S/N ratio of a reproduced picture can be maintained at a high level, by monitoring the encoding mode of a block, determining whether a scene change occurred, and switching the reference picture for motion-compensation prediction.
When a scene change is detected, the picture immediately after the scene change makes an intra picture immediately information amount is compressed, an encoding error is caused only by the difference in this field (or frame), thereby reducing the effect on the motion-compensation prediction process of the next field (or frame). Namely, when motion-compensation prediction is performed before and after a scene change, the amount of information generated in the field (or frame) is increased, such that even when the information is compressed, the effect of an encoding error increase. Therefore, when the same amount of information as that of the generated information is encoded as an intra-field (or intra-frame), higher subjective evaluation can be obtained. In the further high-efficiency encoder of the invention, when a scene change occurs in a motion-compensation prediction process unit, the field (or frame) immediately after the scene change is encoded as an intra-field (or intra-frame). Accordingly, even when the information amount of the field (or frame) is compressed to the level of a prediction-field (or prediction frame), by performing encoding of the field as an intra-field (or intra-frame), the deterioration of the image quality of a picture can be suppressed more efficiently.
In the video information recording/reproducing apparatus of the invention, input signals of n fields (or n frames) are collected into one recording unit block, and recorded in tracks of a predetermined number which is calculated from the amount of information to be recorded and the recording capacity of one track. The compression encoding is performed on blocks which are collected into the recording unit in such a manner that at least one intra-field (or intra-frame) is included in the unit. Input television signals of n fields (or n frames) are subjected as one recording unit block to the compression encoding by a high-efficiency encoder. The compression-encoded television signals of n fields (or n frames) are divided to be recorded in recording areas of m tracks. The reproduced signals of m tracks are restored to television signals of n fields (or n frames) by a high-efficiency decoder.
The above and further objects and features of the invention will more fully be apparent from the following detailed description with accompanying drawings.
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings.
As a predictive method used in such a circuit block, for example, the method shown in
Then, the operation will be described. Irrespective of the field (an intra-field or a prediction-field), digital video signals input through the digital video signal input terminal 1 are blocked by the blocking circuit 2 in the unit of, for example, 8 pixels×8 lines. The motion-compensation predicting circuit 30 performs motion-compensation prediction on input blocks which are output from the blocking circuit 2 in the case of a prediction-field, while using as a reference picture the reproduced video data of an intra-field stored in the video memory 38. The motion-compensation predicting circuit 30 performs motion estimation while setting the search range of motion estimation at, for example, 16 pixels×16 lines, to obtain a motion vector, and further obtains the difference signal between the reference picture and the input image in accordance with the motion vector obtained in the motion estimation. The difference signal is output together with the motion vector to determiner 31. The determiner 31 obtains the sum of absolute values of components of each of the input blocks output from the blocking circuit 2 and the difference blocks output from the motion-compensation predicting circuit 30. When the input blocks are indicated by I(i, j) (i, j=1 to 8), the sum of their absolute values is indicated by Is, the difference blocks are indicated by P(i, j) (i, j=1 to 8), and the sum of their absolute values is indicated by Ps, Is and Ps can be expressed by the following expressions:
When Ps<Is, it is determined that the information amount of the difference blocks is smaller than that of the input signal blocks, and then the difference blocks are output together with the motion vectors to first switch 32. In contrast, when Ps≧Is, it is determined that the information amount of the input signal blocks is smaller than that of the difference signal blocks, and then the input blocks are output together with a forced intra signal indicating that the block is the forced intra block, in place of the motion vectors, to the first switch 32.
The first switch 32 selects the output of the blocking circuit 2 in the case of an intra mode, and the output of the determiner 31 in the case of a prediction mode, and supplies the selected output to the orthogonal transform circuit 33. The orthogonal transform circuit 33 performs, for example, the two-dimensional DCT on each of the 8×8 blocks which are input thereinto. The quantizing circuit 34 performs variable length coding and quantizes the orthogonal transform coefficients output from the orthogonal transform circuit 33. Furthermore, in the case of a prediction mode, quantizing circuit 34 quantizes also the motion vectors or the forced intra signals in addition to the orthogonal transform coefficients, and outputs them together to a transmission path 11. On the other hand, in order to produce reference data for the motion-compensation prediction, only in the case of the intra-field, the second switch 35 outputs the orthogonal transform coefficients which have been quantized by the quantizing circuit 34, to the inverse quantizing circuit 36. The inverse quantizing circuit 36 inverse-quantizes and performs variable length decoding the data which have been performed variable length coding by the quantizing circuit 34, and the decoded data are output to inverse orthogonal transform circuit 37. The inverse orthogonal transform circuit 37 performs, for example, the inverse two-dimensional DCT on the data to reconstruct the blocks of the intra-field. The blocks of the intra-field which have been reconstructed by the inverse orthogonal transform circuit 37 are stored in the video memory 38. The video memory 38 stores the reconstructed pictures of the intra-field for one field, as the reference data in the case of the motion-compensation prediction, and outputs the reference pictures within the motion vector detection range to motion-compensation predicting circuit 30.
In the embodiment described above, a block for the orthogonal transform has the size of 8 pixels×8 lines. This size is not necessarily 8 pixels×8 lines, and the block may have the size of n pixels×m lines. Similarly, it is not necessary to set, the search range for motion vectors Lo 0.1.6 pixels×16 lines, and the search range may be set to k pixels Xs lines (k ^n, s<£ m). In the above embodiment, the predictive encoding is completed every four fields. It is not necessary to complete the predictive encoding every four fields, and the predictive encoding may be completed every arbitrary number of fields. Furthermore, it is not necessary to complete the predictive encoding every arbitrary number of fields, and predictive encoding may be completed every arbitrary number of frames. According to the above embodiment, in the determiner 31, the output having the smaller sum of absolute values is selected from the outputs of the blocking circuit 2 and motion-compensation predicting circuit 30, to be output to the first switch 32. Alternatively, without performing the motion-compensation determination, only the output of the motion-compensation predicting circuit 30 may be output to the first switch 32.
According to Embodiment 1 described above, in the determiner 31, the output having the smaller sum of absolute values is selected from the outputs of the blocking circuit 2 and motion-compensation predicting circuit 30, to be output to the first switch 32. In a field where determiner 31 selects the forced intra mode more frequently than the prediction mode, it may be judged that a scene change has occurred in the field, and assuming that the whole field is in the intra mode encoding may be performed. An embodiment constructed so as to perform this operation is Embodiment 2 described below.
Then, the operation will be described. In the process between the digital video signal input terminal 1 and motion-compensation predicting circuit 30, operation is carried out in the same manner as those in the first embodiment, and therefore the description is omitted. The determiner 40 selects the one having the smaller sum of absolute values of components, from input blocks output from the blocking circuit 2 and difference blocks output from the motion-compensation predicting circuit 30, and outputs it. When the output from the motion-compensation predicting circuit 30 is selected, the determiner 40 outputs motion vectors and blocks of difference signal. When the output from the blocking circuit 2 is selected, motion vectors are output together with a signal indicative of a forced intra-block. When the number of forced intra-blocks in one field readies or exceeds a value n, determiner 40 judges that a scene change has occurred, and outputs a control signal so that the whole current field is to be encoded by the intra mode.
The output of the determiner 40 is stored as data of the prediction mode in the second field memory 42, and, after data for one field have been stored, it is output to the first switch 43. In contrast, the output of the blocking circuit 2 is stored as data of the intra mode in the first field memory 41, and, after data for one field have been stored, it is output to the first switch 43. In the intra mode and when the determiner 40 determines to force the intra mode on the field, the first switch 43 selects the output of the first field memory 41, and, in another case, selects the output of second field memory 42. The output of the first switch 43 is supplied to the orthogonal transform circuit 33. In the succeeding process from the orthogonal transform circuit 33 to the video memory 38, the operation, is carried out in the same manner as those in the first embodiment, and therefore the description is omitted. However, when it is determined by the determiner 40 that a scene change occurs, it is required to update also the contents of the video memory 38. In this case, therefore, the second switch 35 operates, in the same manner as in the intra mode, to supply the output of the quantizing circuit 34 to the inverse quantizing circuit 36.
Hereinafter, examples in which encoding and decoding are respectively performed by three kinds of predictive encoding methods shown in
In Embodiment 2 described above, the intra mode is created every n fields and the succeeding (n−1) fields are prediction-encoded from an intra-field, and, when a scene change occurs, the intra-field is forcedly created and the remaining fields are predicted from the forced intra mode. It is not necessary that an intra-field exists every n fields. Alternatively, when a forced intra-field is created, the succeeding (n−1) fields which comes after the forced intra-field may be prediction-encoded. In the above-described embodiment, the predictive encoding is performed in the unit of field. It is not necessary to perform predictive encoding in the unit of field, and predictive encoding may be performed in the unit of frame.
As described above, since the high-efficiency encoders of Embodiments 1 and 2 create an intra-field every n fields and in the other fields, perform motion-compensation prediction using this intra-field as a reference picture, the hardware size including a calculation circuit for obtaining motion vectors can be reduced.
Then, the operation will be described. Irrespective of an intra-field or a prediction-field, input digital video signals are by the blocking circuit 2 segmented into input blocks a unit of which consists of m [pixels]×n [lines]. In order to obtain a difference block, subtracter 3 calculates the difference in the unit of pixel between an input block and a prediction block. In this way, an input block and a difference block are supplied to first switch 7. In order to calculate the power, the difference block is supplied also to difference power calculator 4 and the difference power is calculated. In order to calculate the AC power, the input block is supplied also to the original power calculator 5 and the original power is calculated. The outputs of the difference power calculator 4 and original power calculator 5 are supplied to the determiner 6 which in turn selects the one having the smaller power from the two powers outputting the result to the first switch 7 as the mode signal. More specifically, when the difference power is smaller than the original AC power, the prediction mode signal is output so that the first switch 7 is set to the prediction mode to output the difference block as an encoded block. When the original power is smaller than the difference power, the intra mode signal is output so that first switch 7 is set to the intra mode to output the input block as an encoded block.
The mode signal from determiner 6 is input to the mode counter 50. Since the input mode signal is generated for every block of a prediction-field, the mode counter 50 counts the number of blocks selecting the intra mode or prediction mode, from the blocks of one field, and outputs the number of blocks selecting an intra mode or prediction mode to the direction switching circuit 51. The direction switching circuit 51 compares a predetermined number of blocks (which is smaller than the total number of blocks of one field) with the number of blocks selecting an intra mode which number has been input from mode counter 50, and outputs a reference picture switching signal to video memory 52. When the predetermined number of blocks is greater (or smaller) than the number of blocks selecting an intra mode (or a prediction mode), direction switching circuit 51 outputs to video memory 52 a reference picture switching signal so that the reference picture is not to be switched. When the number of predetermined blocks is smaller (or greater) than the number of blocks selecting an intra mode (or a prediction mode), the direction switching circuit 51 outputs to the video memory 52 a reference picture switching signal so that the reference picture is to be switched.
The first switch 7 outputs an input block or a difference block as an encoded block, depending on the mode signal determined by the determiner 6. In this case, when the input block is the intra-field, the first switch 7 surely outputs the input blocks as encoded blocks. The encoded blocks enter the DCT circuit 8 to be converted into DCT coefficients. The DCT coefficients are subjected to the weighting and threshold processes in the quantizing circuit 9 to be quantized to predetermined bit numbers each corresponding to the coefficients. The DCT coefficients which have been quantized to the respective predetermined bit numbers are converted by the first encoder 10 into codes suitable for transmission path 11 and then output to transmission path 11.
The DCT coefficients on which weighting and threshold processes and quantization have been performed by the quantizing circuit 9 also enter the local decoding loop 20, and are subjected to the inverse weighting and inverse quantizing processes in the inverse quantizing circuit 12. Then, the DCT coefficients on which inverse weighting and inverse quantizing processes have been performed in the local decoding loop 20 are converted into a decoded block by inverse DCT circuit 13. The adder 14 adds the decoded block to a prediction block in the unit of pixel. This prediction block is the same as that used in the subtracter 3. The result of addition by the adder 14 is written as an output block in a predetermined address of the video memory 52. Video memory 52 switches the reference picture in response to the reference picture switching signal from the direction switching circuit 51, and outputs the motion estimation search range to the MC circuit 16. The size of the block in the motion estimation search range is, for example, i [pixels]×j [lines] (where i≧m, j≧n). A block in the motion estimation search range output from the video memory 52, and an input block from the blocking circuit 2 are input into the MC circuit 16. The MC circuit 16 performs the motion estimation on each block to extract motion vectors of the input block.
The motion vectors extracted by the motion estimation in the MC circuit 16 are input into the MIX circuit 17. The MIX circuit 17 combines the motion vectors from the MC circuit 18 with the mode signal determined by the determiner 6. The motion vectors and mode signal which have been combined with each other in the MIX circuit 17 are converted by the second encoder 18 into codes suitable for the transmission path 11, and then output together with the corresponding encoded blocks to the transmission path 11. The prediction block is output from the MC circuit 16 in the form of a block which is blocked into the size (m [pixels]×n [lines]) equal to that of the input block from the motion estimation range. This prediction block is supplied to the second switch 19, and output from the respective output terminal of the switch, in accordance with the field of the input block the currently being processed and the mode signal of the decoded block. Namely, the prediction block is output from one of the output terminals of the second switch 19 to the subtracter 3 in accordance with the processed field, and from the other output terminal in accordance with the mode signal of the current decoded block and the processed field.
According to the present invention, in the case that a scene change occurs in the unit of frame when the predictive method shown in
The operation in Embodiment 3 is summarized below with reference to flowcharts in
Firstly, the field number fn indicative of the field in the motion-compensation prediction process unit is set to be 0 (step S101). Referring to
Then, the field number fn is checked to determine whether it is 0 or whether the field is the first field in the motion-compensation prediction process unit and an intra-field (step S102). If the field number fn is 0 (fn=0), this field is processed as an intra-field (step S103). In contrast, if the field number fn is not 0 (fn≠0), this field is processed as a prediction-field. These processes will be described in detail later. After each field is processed, the field number fn is incremented so as to indicate the next field (step S105). In an actual hardware, such a field number can be controlled by a microcomputer signal or the like.
It is judged whether the field number fn indicating the next field is a number indicating a field within the motion-compensation prediction process unit (step S106). If the field number fn is a number not indicating a field within the motion-compensation prediction process unit, for example, as in
When a series of the motion-compensation prediction process units have been completed, it is judged whether all of the required fields have been processed (step S107). This judgment can be done by, for example, checking the operation of an end switch of the high-efficiency encoder. If the next field is to be processed, in order to encode the next motion-compensation prediction process unit, the variables are initialized, and the process is repeated. If the operation of the high-efficiency encoder has been completed, the encoding is ended.
The intra-field process will be described with reference to the flowchart in
The prediction-field process will be described with reference to the flowchart in
Then, a variable COUNT for counting the number of blocks which select an intra mode in one field to be processed is set to be 0 (step S304). The variable COUNT will be described in detail later. An input picture is segmented into the predetermined size of m [pixels]×n [lines] in the processed field (step S305). The blocks segmented into the size of m×n are subjected to motion-compensation prediction (step S306). Using the reference picture set in step S302 or S303, the difference in the unit of pixel between a predetermined area of a past image and a newly divided block is input as a difference block into difference power calculator 4, and a difference power P1 is calculated (step S307). Namely, by using such preset reference picture, the information amount generated by motion-compensation prediction can be reduced. Then, the newly segmented block is input into original power calculator 5 to calculate an original AC power P2 (step S308).
The calculated powers P1 and P2 are compared in magnitude with each other (step S309). When the difference power P1 is smaller than the original AC power P2, the difference block (the difference value of the block subjected to motion-compensation prediction) is selected (step S310). When the difference power P1 is greater than the original AC power P2, the input block (the original which is still in block form is selected (step S311), and the number of times when an input block is selected as an encoded block or the number of blocks in one field which are to be processed as an intra mode is counted (step S312). A variable that functions as a counter in this step is the COUNT which has been set to be 0 in step S304. This counter is surely set to be 0 when the process in the unit of field starts, and counts the number of blocks which select an intra mode in the processing of one field.
Each selected block is subjected to the orthogonal transform (step S313), and quantized to a predetermined bit number which is set for each sequence (step S314). In an orthogonal transform such as DCT, a quantization is performed in such a manner that a larger bit number is assigned to a DC and low-order sequences of AC, and a smaller bit number is assigned to a high-order sequence of AC. The quantized data are converted into codes suitable for transmission (step S315), and the encoded data are transmitted (step S316). For example, the number of processed blocks is counted to judge whether the process of one field has been completed (step S317). If the process of one field has not yet been completed, the process for the next block is pursued.
If the process of one field has been completed, the number of input blocks which have been processed as encoded blocks in the process of that one field or the number of blocks which have selected an intra mode is compared with a preset threshold TH (step S318). The threshold TH is a predetermined number which is less than the number of blocks in one field. When the total number of blocks in one field is 2,700, for example, the threshold TH is set to be 1,000 which is less than 2,700. If the variable COUNT indicating the number times when input blocks being selected as encoded blocks is smaller than the threshold TH, there is no scene change between the field (field number fn) which has been just processed and the reference picture which has been used in motion-compensation prediction of that field, and the reference picture switch flag Rfn is set to be 0 (Rfn=0) so that the reference picture in a normal position is used as the reference picture for the motion-compensation of the next field (field number fn+1) (step S319). If the variable COUNT indicating the number of times when input blocks being selected as encoded blocks is greater than the threshold TH, there is a scene change between the field (field number fn) which has been just processed and the reference picture which has been used in motion-compensation prediction of that field, and the reference picture for the motion-compensation of the next field (field number fn+1) is switched from the reference picture in a normal position to a field in a position different from the position taken till then, for example, the field which has been just processed and is positioned in a place where a reference picture has not existed till then. For that purpose, the reference picture switch flag Rfn is set to be 1 (Rfn=1) (step S320). In this way, the reference picture switch flag Rfn is set and the prediction-field process is ended.
In Embodiment 3, in order to select an encoded block from a difference block and an input block, their powers are calculated and compared with each other, and the number of blocks selecting an intra mode are counted.
According to Embodiment 4, in order to select an encoded block from a difference block and an input block, the entropy in each block is calculated, and the entropy of the difference block is compared with that of the input block in the same manner as in Embodiment 3 by the determiner 6, to determine which block is to be selected as an encoded block.
According to Embodiment 5, in order to select an encoded block from a difference block and an input block, adding of absolute values of pixels is performed in each block, the rth power of the sum of absolute values of the input block and that of difference block are calculated, and the rth power of the sum of absolute values of the difference block is compared with that of the input block in the same manner as in Embodiment 3 by the determiner 6, to determine which block is to be selected as an encoded block.
In Embodiment 3, the determiner 6 compares the power of an input block with that of a difference block. According to Embodiment 6, when the power of an input block is to be compared with that of a difference block, at least one of the powers of the input and difference blocks is provided with an offset, and then the two powers are compared with each other. For example, the power of the input block is provided with a positive offset, and then compared with the power of the difference block. When there is not a great difference in power between the input and difference blocks, this configuration allows the number of blocks selecting the difference power to be increased, thereby preventing an intra mode from being excessively created.
According to Embodiment 7, when the entropy of a difference block is compared with that of an input block in a similar manner as in Embodiment 4, at least one of the entropy of the input and the entropy of the difference block is provided with an offset, and then the two values are compared with each other. For example, the entropy of the input block is provided with a positive offset, and then compared with the entropy of the difference block. When there is not a great difference between the entropy of the input block and the entropy of difference block, this configuration allows the number of blocks selecting difference power to be increased, thereby preventing an intra mode from being excessively created.
According to Embodiment 8, when the sum of absolute values of a difference block is compared with that of an input block in a similar manner as in Embodiment 5, at least one of the rth power of the sum of absolute values of the input block and that of the difference block is provided with an offset, and then the two values are compared with each other. For example, the rth power of the sum of absolute values of the input block is provided with a positive offset, and then compared with the rth power of the sum of absolute values of the difference block. When there is not a difference greater than the offset between the rth power of the sum of absolute values of the input and that of the difference block, this configuration allows the number of blocks selecting the difference power to be increased, thereby preventing an intra mode from being excessively created.
In Embodiment 3, the mode counter 50 counts the number of all blocks selecting an intra mode among blocks for one field. In Embodiment 9, blocks for one field are not counted, but, at the time a mode signal in a predetermined number of blocks during one field is determined, the ratio of the number of blocks selecting an intra mode to the total number of blocks or the number of the blocks in which a mode signal has been determined is supplied to the direction switching circuit 51. Based on this ratio, a reference picture switching signal is output from the direction switching circuit 51. This configuration allows the reference picture for the next field to be determined even when encoding of all blocks for one field has not yet completed.
Embodiment 10 will be described with reference to
Then, the operation will be described. It is assumed that motion-compensation prediction is performed, for example, as shown in
An input block which is obtained only by blocking an original output from the video memory 60, and a difference block that is a difference between the input block and a prediction block which has been subjected to motion-compensation prediction by the subtracter 3 are input to the first switch 62. The input block and the difference block are respectively input to the original power calculator 5 and the difference power calculator 4 in order to obtain the power of each block. The original power calculator 5 calculates the AC power of the input block, and the difference power calculator 4 calculates the power of the difference block. The calculated AC power of the input block and the calculated power of the difference block are input to the determiner 6. When the power of the difference block is smaller than that of the input block, the determiner 6 outputs a prediction mode signal, and, when the power of the input block is smaller than that of the difference block, the determiner 6 outputs an intra mode signal. These signals are supplied as a mode signals to the first switch 62, the MIX circuit 63 and the second switch 64.
The first switch 62 to which the input and difference blocks are input outputs either of the blocks as an encoded block. For that purpose, the first switch 62 receives the scene change detection signal from the SC detection circuit 61 and also the mode signal from the determiner 6, to determine the switch mode, and outputs either of the input and difference blocks as an encoded block. The switching states at this time are shown in
The encoded block which has been selected and output by the first switch 62 is subjected to two-dimensional orthogonal transform by the DCT circuit 8. The orthogonal-transformed data is subjected to the weighting and threshold processes or the like in the quantizing circuit 9 to be quantized to a predetermined bit number in the respective sequence. The data quantized by the quantizing circuit 9 are converted by the first encoder 10 into codes suitable for the transmission path 11 and then output to the transmission path 11. The data quantized by the quantizing circuit 9 are input also to the local decoding loop 20 so that motion-compensation prediction is performed. The data input to the local decoding loop 20 are subjected to inverse quantizing and inverse weighting processes in inverse quantizing circuit 12, and then subjected to the inverse orthogonal transform by the inverse DCT circuit 13. A decoded block which is an output of inverse DCT circuit 13 is added in the unit of pixel to the prediction block in the adder 14 to become a reproduced image. The prediction block used in this process is identical with that used in the subtracter 3. The block which has became a reproduced image in the adder 14 is written in a predetermined address of the video memory 15.
The memory size of the video memory 15 depends on the type of the employed predictive method. In this embodiment, it is assumed that the video memory 15 consists of a plurality of field memories, and that output blocks reconstructed by the local decoding loop 20 are stored in a predetermined address. These stored images are used as data of the search range for motion-compensation prediction. A block which is segmented from an image reconstructed from past output blocks and is in a motion estimation search range is output from the video memory 15 to the MC circuit 16. The size of the block of the motion estimation search range is i [pixels]×j [lines] (where i≧m, j≧n, and i and j are positive integers). Data in the search range for motion-compensation prediction from video memory 15 and an input block from video memory 60 are input to the MC circuit 16 as reference data, thereby motion vectors being extracted.
The motion vectors extracted by the MC circuit 16 are input to the MIX circuit 63, and combined therein with the mode signal determined by the determiner 6 and the SC detection signal from the SC detection circuit 61. The combined signals are converted by the second encoder 18 into codes suitable for the transmission path 11, and then output together with the corresponding encoded block to the transmission path 11. The MC circuit 16 outputs blocked signals which are segmented from the search range in the size (m [pixels]×n [lines]) equal to that of the input block, as a prediction block. The prediction block output from the MC circuit 16 is produced from past video information. The prediction block is supplied to the second switch 64, and output in accordance with the currently processed field, the mode signal of the decoded block and the SC detection signal from the SC detection circuit 61. Namely, the prediction block is output from one of the output terminals of the second switch 64 to the subtracter 3 in accordance with the processed field and the SC detection signal, and from the other output terminal in accordance with the mode signal of the current decoded block, the SC detection signal and the processed field.
The motion-compensation prediction process is shown in
The operation in Embodiment 10 will be summarized with reference to flowcharts of
At first, the field number indicative of a field in a motion-compensation prediction process unit is set to be 0 (step S401). This setting of the field number is the same as that in Embodiment 3. Since the motion-compensation process has been just started, the field to be initially processed is surely the first field in the motion-compensation prediction process unit and an intra-field, and therefore the field number fn is set in step S401 to be 0. A scene change detection flag Cfn which functions as a flag for judging whether a scene change is present is set in step S401 to be 0 for the initialization.
Then, the characteristics of an input image are compared with those of a past image by a certain parameter to detect the presence of a scene change (step S402). For example, the presence of a scene change is detected by comparing the variance of values of pixels in some predetermined areas of the past image with the variance of values of pixels in some predetermined areas of the currently processed image. When a scene change is detected, the scene change detection flag Cfn is set to be 1 (Cfn=1), and, when a scene change is not detected, the scene change detection flag Cfn is set to be 0 (Cfn=0).
Then, the field number fn is checked to judge whether it is 0 or the field is the first field in the motion-compensation prediction process unit and an intra-field (step S403). If the field number fn is 0 (fn=0), this field is processed as an intra-field (step S405). In contrast, if the field number fn is not 0 (fn≠0), the process proceeds to next step S404. It is judged whether the scene change detection flag Cfn is 0 (Cfn=0) or there is a scene change between the processed field and the reference picture required for encoding the field with motion-compensation prediction (step S404). If the flag Cfn is 0 (Cfn=0}, there is no scene change, and the field to be processed is processed as a prediction-field (step S406). If the flag Cfn is 1 (Cfn=1), there exists a scene change, and therefore the field to be processed is processed as an infrafield (step S405). Therefore, even in the case that a field is in the motion-compensation prediction process unit and the field number is not 0, when a scene change is detected and the flag Cfn is 1, the field is processed as an intra-field.
After each field is processed, the field number fn is incremented so as to indicate the next field (step S407). In an actual hardware, such a field number can be controlled by a signal from a microcomputer or the like.
It is then judged whether the field number fn indicating the next field is a number indicating a field within the motion-compensation prediction process unit (step S408). If the field number fn is a number not indicating a field within the motion-compensation prediction process unit, for example, in
Next, the prediction-field process (step S406 in
The calculated powers P1 and P2 are compared in magnitude with each other (step S455). When the difference power P1 is smaller than the original AC power P2, the difference block (the difference value of the block subjected to motion-compensation prediction) is selected (step S456). When the difference power P1 is greater than the original AC power P2, the input block (the original which has been blocked and not subjected to any further process) is selected (step S457). Each selected block is subjected to the orthogonal transform (step S458), and the quantized to a predetermined bit number which is set for each sequence (step S459). In an orthogonal transform such as DCT, for example, a quantization is performed in such a manner that a larger bit number is assigned to a DC and low-order sequences of AC, and a smaller bit number is assigned to a high-order sequence of AC. The quantized data are converted into codes suitable for transmission (step S460), and the encoded data are transmitted (step S461). It is judged by, for example, counting the number of processed blocks whether the process of one field has been completed (step S462). If the process of one field has not yet been completed, the process for the next block is pursued. If the process of one field has been completed, the prediction field process is ended.
According to Embodiment 10, when a scene change occurs in a motion-compensation prediction step as shown in
In Embodiment 10, even when a scene change occurs in a step of a motion-compensation prediction process and the field immediately after the scene change is set to be an intra-field, the time-constraint length in the motion-compensation prediction processing step is fixed for four fields. Namely, an intra-field surely appears every four fields once the motion-compensation prediction process starts, and, when a scene change occurs, an intra-field exists also in the motion-compensation processing step. This is a configuration in which a prediction-field is replaced with an intra-field.
According to Embodiment 11, when a scene change occurs as shown in
According to Embodiment 11, the image immediately after the scene change is set to be an intra-field, thereby improving a subjective appreciation of the image. When the frequency of the occurrence of scene changes is longer than the time-constraint length in the motion-compensation prediction process unit and is low in level, the number of fields of an intra-field is smaller than that in Embodiment 3 so that the information amount can be reduced.
In Embodiments 10 and 11, a process is performed while setting a field (or frame), in which a scene change is detected, to be an intra-field (or intra-frame). Alternatively, without setting as an intra-field (or intra-frame), the reference picture of the field (or frame) may be set to be intra-field (or intra-frame) belonging to the next motion-compensation prediction process unit.
Embodiment 12 will be described with reference to
In Embodiment 12, the motion-compensation, prediction process unit is sometimes longer than the usual one. According to Embodiment 13, P fields (or P frames) which begin with a field (or frame) wherein a scene change is detected and have the total length corresponding to the length of a usual motion-compensation prediction process unit are subjected to motion-compensation prediction.
Embodiment 13 will be described with reference to
In Embodiments 3 to 13, a switching is explained with reference to a scene change. A reference picture may be switched in accordance with the number of intra modes which have been forcedly generated in blocks. Accordingly even in a picture wherein many forcible intra modes appear, or in a case that an object which has not existed in one field before appears suddenly in the current field or an object which has existed in one field before disappears suddenly from the current field, a reference picture can be switched by a similar method.
In Embodiments 3 to 13, for example, a process in which motion-compensation prediction as shown in
In Embodiments 3 to 15, the motion-compensation prediction process is performed in the unit of four fields. The number of fields which may be used as the unit is not necessarily four. The process may be performed in the unit of an arbitrary number of fields on which the motion-compensation prediction process can be performed.
According to Embodiments 3 to 16, without largely increasing the memory amount by additionally providing a hardware as described above, even when a scene change occurs in the motion-compensation prediction process unit, a reference picture is switched from the originally set reference picture so as to minimize the influence caused by the scene change, etc., the image immediately after the scene change is set to be a reference picture for motion-compensation prediction, and, after the detection of the scene change, fields before the scene change are not used as the reference picture for motion-compensation prediction, whereby transmission can be done while suppressing the increase of the code amount due to the scene change to a minimum, and without deteriorating the image quality.
When motion-compensation prediction is performed before and after a scene change, the information amount of the predicted picture is usually increased. Therefore, by processing the field as an intra picture with an information amount equal to that information amount, a subjective appreciation of the picture can be improved. While setting a picture immediately after a scene change as an intra picture by detecting the scene change, encoding between fields or frames can be performed so that a subjective appreciation of the picture immediately after the scene change can be improved. When a scene change occurs, the picture immediately after the scene change is handled as an intra picture, and motion-compensation prediction is performed with the intra picture being the first picture, whereby the number of generated intra pictures can be reduced and the amount of generated information can be reduced.
The recording operation will be described. A video signal input to the A/D converter 101 is converted into a digital video signal, and output to the high-efficiency encoder 110. The high-efficiency encoder 110 performs the reduction of redundancy using auto-correlation of the video information, the human visual characteristics and the bias of the data generation frequency, to compress the information (its detail will be described later). The output of the high-efficiency encoder 110 is supplied to the error-correction encoder 102 in which error-correcting codes for correcting transmission errors are added to it. In this process, codes having a high error correction capability and a small information amount to be added are used, in order to perform a high-density recording and because even, a small error in compressed information exerts an influence over a wide range. The data to which error-correcting codes have been added are modulated by the modulator 103 to a signal suitable for magnetic heads 106 and the magnetic tape 105. The modulator 103 also performs other operations such as suppression of DC and low-frequency components for the azimuth recording, and the addition of a tracking signal which assists the trace function of the magnetic heads 106. The record signal which has been modulated by the modulator 103 is recorded through the magnetic heads 106 on the magnetic tape 105. The magnetic heads 106 are mounted on the rotary head drum 104 so as to be rotated by the rotation of the drum 104. The so-called helical scanning recording is conducted on the magnetic tape 105.
Next, the reproduction operation will be described. The signal which has been recorded by the helical scanning on the magnetic tape 105 is picked up by the magnetic heads 106 mounted on the rotary head drum 104, and then demodulated by the demodulator 107. The demodulated signal is subjected to the error detection and error correction by the error-correction decoder 108. The error-corrected data are expanded by the high-efficiency decoder 111 to be changed from the compressed codes to the original digital video signal. The reconstructed digital video signal is converted to an analog video signal by the D/A converter 109, and then output.
The operation of the high-efficiency encoder 110 will be described. The initial field of a recording unit block is encoded as an intra-field which does not use the inter-plane prediction. Since the first switch 302 selects the upper contact, an input digital video signal is subjected to orthogonal transform by the DCT circuit 303. The transformed data are quantized by the quantizing circuit 304, and encoded by the variable-length encoder 305 to a variable-length code such as a Huffman code, to be output. At the same time, the quantized data are inverse-quantized by the inverse quantizing circuit 306, and then supplied to the inverse DCT circuit 307. In the inverse DCT circuit 307, the orthogonal-transformed data are inverted to the original video data and then output to the adder 308. In the intra-field, the second switch 311 also selects the upper contact so that the one input of the adder 308 is zero. Therefore, the output of the inverse DCT circuit 307 is supplied as it is to the video memory 309 to be stored thereinto.
In encoding of the next field, interplane prediction is used. In interplane prediction, both the first and second switches 302 and 311 select their lower terminal. An input digital video signal enters the subtracter 301 and motion-compensation circuit 310. The motion-compensation circuit 310 compares the stored picture with the input picture, and outputs motion vectors of the input picture and a prediction picture to be used in the predictive encoding. The subtracter 301 calculates the difference between the input picture and the prediction picture, and output it as a prediction difference signal to the DCT circuit 303. Compared with a raw video signal, a prediction difference signal has a smaller information amount as the prediction accuracy becomes higher. For example, a prediction difference signal for an entirely still picture is zero. In the same manner as the initial field, the data input to the DCT circuit 303 are subjected by the DCT circuit 303 and quantizing circuit 304 to the orthogonal transform and quantization, and then converted into variable-length codes by the variable-length encoder 305 to be output. On the other hand, the quantized data are supplied through the inverse quantizing circuit 306 to the inverse DCT circuit 307 to be subjected to the inverse quantization and inverse orthogonal transform, and then sent to the adder 308. Since the prediction picture used in the process of obtaining the prediction difference is supplied to the other input terminal of adder 308, the output of the adder 308 is the same as the input picture. The output of the adder 308 is stored into the video memory 309. In the same manner as described above, the process of encoding n fields is pursued.
In Embodiment 17, data of four fields are recorded in ten tracks. It is not necessary to always record data in ten tracks, and data may be recorded in eight or six tracks depending on the amount of information to be recorded.
As described above, in Embodiments 17 and 18, since signals of a plurality of fields or frames are collected in one recording unit to be recorded in a predetermined number of tracks, all of the recorded fields can be reconstructed by performing reproduction processing on a predetermined number of tracks. Therefore, the embodiments can cope with the special reproduction and edition required in a VTR. Since the number of tracks to be used in the recording is selected depending on the amount of information to be recorded, there is no wasted track, thereby recording and reproduction can be performed for a long period. Furthermore, since it is not necessary to control information to be recorded so as to match the recording capacity of each track, there is no useless portion in each track, with the result being that recording can be performed efficiently. Moreover, since an intra-picture which does not use interplane prediction surely exists in each recording unit, a reconstructed picture can be easily obtained even in a special reproduction such as a speed search, and the amount of information to be recorded can be reduced, as compared with a prediction picture which uses the interplane prediction.
As this invention may be embodied in several forms without departing from the spirit of essential characteristics thereof, the present embodiment is therefore illustrative and not restrictive, since the scope of the invention is defined by the appended claims rather than by the description preceding them, and all changes that fall within metes and bounds of the claims, or equivalence of such metes and bounds thereof are therefore intended to be embraced by the claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 4-013719 | Jan 1992 | JP | national |
| 4-037599 | Feb 1992 | JP | national |
| 4-037821 | Feb 1992 | JP | national |
| 4-043075 | Feb 1992 | JP | national |
This application is a divisional of co-pending application Ser. No. 10/372,212 filed on Feb. 25, 2003, which is a divisional of application Ser. No. 09/271,458, filed on Mar. 18, 1999, now U.S. Pat. No. 6,870,884 B1, which is a divisional of application Ser. No. 09/113,287, filed Jul. 10, 1998, now U.S. Pat. No. 5,909,252, which is a divisional of application Ser. No. 08/559,488, filed Nov. 15, 1995, now U.S. Pat. No. 5,841,474, which is a divisional of application Ser. No. 08/011,243, filed on Jan. 29, 1993, now U.S. Pat. No. 5,479,264, the entire contents of which are hereby incorporated by reference and for which priority is claimed under 35 U.S.C. § 120; and this application claims priority of Application No. 4-013719, 4-037599, 4-037821, and 4-043075 filed in Japan on Jan. 29, 1992; Feb. 25, 1992; Feb. 25, 1992 and Feb. 28, 1992 under 35 U.S.C. § 119.
| Number | Name | Date | Kind |
|---|---|---|---|
| 4520401 | Takahashi et al. | May 1985 | A |
| 4651206 | Ohki | Mar 1987 | A |
| 4689673 | Ohki et al. | Aug 1987 | A |
| 4827338 | Gerard | May 1989 | A |
| 4837618 | Hatori et al. | Jun 1989 | A |
| 5047852 | Hanyu et al. | Sep 1991 | A |
| 5089888 | Zdepski et al. | Feb 1992 | A |
| 5089889 | Sugiyama | Feb 1992 | A |
| 5091782 | Krause et al. | Feb 1992 | A |
| 5132792 | Yonemitsu et al. | Jul 1992 | A |
| 5144423 | Knauer et al. | Sep 1992 | A |
| 5146325 | Ng | Sep 1992 | A |
| 5227878 | Puri et al. | Jul 1993 | A |
| 5231484 | Gonzales et al. | Jul 1993 | A |
| 5260783 | Dixit | Nov 1993 | A |
| 5309292 | Takakura | May 1994 | A |
| 5528238 | Nickerson | Jun 1996 | A |
| 5565998 | Coombs et al. | Oct 1996 | A |
| 5805762 | Boyce et al. | Sep 1998 | A |
| Number | Date | Country |
|---|---|---|
| 3831277 | Mar 1990 | DE |
| 0267581 | May 1988 | EP |
| 0289960 | Nov 1988 | EP |
| 0353758 | Feb 1990 | EP |
| 0434427 | Jun 1991 | EP |
| 0493136 | Jul 1992 | EP |
| 0536630 | Apr 1993 | EP |
| 02-174388 | Jul 1990 | JP |
| 02-192378 | Jul 1990 | JP |
| 3139083 | Jun 1991 | JP |
| 04-010788 | Jan 1992 | JP |
| 04-051688 | Feb 1992 | JP |
| 04-051689 | Feb 1992 | JP |
| 4068989 | Mar 1992 | JP |
| 4318785 | Nov 1992 | JP |
| 4326280 | Nov 1992 | JP |
| 5111015 | Apr 1993 | JP |
| WO-8702854 | May 1987 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 20060133487 A1 | Jun 2006 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 10372212 | Feb 2003 | US |
| Child | 11276501 | US | |
| Parent | 09271458 | Mar 1999 | US |
| Child | 10372212 | US | |
| Parent | 09113287 | Jul 1998 | US |
| Child | 09271458 | US | |
| Parent | 08559488 | Nov 1995 | US |
| Child | 09113287 | US | |
| Parent | 08011243 | Jan 1993 | US |
| Child | 08559488 | US |
