The present invention relates to a randomly accessible video information recording medium, a recording device for recording video data on a video information recording medium, the recording method, a reproducing device for reproducing video data from a randomly accessible video information recording medium, and the reproducing method.
A proposed system for writing encoded image data in packets formatted according to the MPEG system on a storage medium to enable trick reproduction of image data by a simple and efficient method without increasing the capacity of the storage medium is to set up an I picture index of packets in which at least part of the I picture data is stored and, during trick reproduction, to read only the packets on which the I picture index is set (for example, Patent Document No. 1).
Patent Document No. 1: Japanese Patent Application Publication No. H9-98430 (pp. 4-10, FIGS. 1-15)
In recent years, MPEG4-AVC (H.264) and other new encoding methods by which even low-bit-rate encoding can provide adequate image quality have come into use. To obtain high image quality with a low bit rate, it is necessary to minimize the number of intra-coded (I) pictures, which require relatively much encoded data.
Because the first frame in a group of pictures (GOP) must be an I picture, reducing the number of I pictures is equivalent to increasing the GOP length. For example, a 1-seg broadcast, which is a television broadcast for mobile terminals, allows up to a five-second GOP length. If the GOP length is increased in this manner, the number of positions that can be designated as access points for random access is greatly reduced, because random access must start from an I picture, which can be decoded without reference to other frames. In a time search etc. of a video image recorded by a user, accordingly, because the only positions designated as accessible points are the I pictures at which GOPs begin, although the user may want to reproduce the image starting from a certain point in time, access to precisely that point of time may be difficult; to the user's inconvenience, it may only be possible to reproduce the image starting from a time offset from the desired point in time. Another problem is that if the GOP length is set to five seconds, for example, in random access to a point within that interval, in the worst case more than two seconds passes before the desired picture is reproduced.
The present invention addresses the above problems, with the object of providing a video information recording medium, a video information recording apparatus and recording method, and a reproducing apparatus and reproducing method with which random access is possible even if the GOP length is increased.
This invention is a video information recording medium on which are recorded video data organized into video units including intra-coded (I) pictures coded intra-frame, predictive coded (P) pictures each including a group of blocks predicted from one frame, and bidirectionally predictive coded (B) pictures each including a group of blocks predicted from two frames, the video information recording medium being randomly accessible and having disposed thereon information indicating a decoding sequence of pictures designated as randomly accessible positions and thus functioning as access points, and attribute information indicating whether a picture functions as an access point or is necessary for decoding of the access point following a given access point.
According to this invention, video reproduction can be carried out smoothly from the point in time desired by the user even if the GOP length is increased, and random access is possible.
a) and
AP, AP1, AP2 access points, 1 navigation data, 2 video data, 3 video file, 4 GOP(k) 100, 120 reproducing device, 101 user interface, 102 CPU, 103 drive, 104 drive controller, 105 work memory, 106 system bus, 107 decoder, 108 buffer memory, 109 display device, 110 picture selector
Embodiments of the invention will now be described with reference to the attached drawings.
The embodiments illustrate mainly the case in which the video information recording medium is an optical disc, but it may be another recording medium such as a hard disc or semiconductor memory.
These embodiments will be described for the case in which a 30 frames per second picture is compressed to digital video data by the MPEG4-AVC (Advanced Video Codec) encoding system. Each frame of digital video data consists of one of three types of coded picture: an I picture, a P picture, or a B picture. An I picture is an intra-coded picture coded within one frame. A P picture is a predictive coded picture; one frame is divided into a group of blocks, each block being predicted from one other frame; that is, a P picture is a group of blocks predicted from one frame. A B picture is also a predictive coded picture in which one frame is divided into a group of blocks, but each block is predicted with reference to two other frames; that is, a B picture is a group of blocks predicted from two frames. A group of pictures (GOP) is structured as a unit comprising at least one I picture followed by one or more P pictures and one or more B pictures, and the video data consist of a plurality of such GOPs. The MPEG4-AVC standard provides no definition of GOPs, but the term GOP will be used herein to denote this concept as applied to MPEG4-AVC, and the following description will assume that the length of a GOP is 1.0 second.
a) and
a) is an example of the MPEG4-AVC coded GOP structure. How the pictures are decoded will be described with reference to
In
As described above, to decode one picture, a temporally preceding picture and/or a following picture is necessary, which poses a significant problem for random access. Consider P2: in order to decode P2, B4 is necessary, and in order to decode B4, B3, and P1 are necessary. In order to decode B3, P1, and B2 are necessary, and in order to decode B2, B1, and P1 are necessary. In order to decode B1, I1, and P1 are necessary. Therefore, if one wants to start reproducing from P2, one finds that P2 can be decoded only after decoding all of pictures I1, B1, B2, P1, B3, and B4. (In MPEG2, which, unlike MPEG4-AVC, does not permit the pictures used for reference and prediction to be selected arbitrarily, P2 could be decoded and reproduced by decoding I1 and P1.) This means that a picture encoded by MPEG4-AVC lacks random accessibility. In low-bit-rate encoding, the GOP length tends to be set high in order to improve the coding efficiency. If, for example, the length of a GOP is about five seconds, then in random access to a picture in the interior of the GOP, in the worst case, more than two seconds passes before the desired picture is reproduced.
Therefore, in an encoding scheme such as MPEG4-AVC which allows arbitrary selection of the pictures used for reference and prediction, many pictures have to be decoded in order to access a picture in the interior of a GOP, and time is necessary before reproduction begins. As a result, problems arise in starting reproduction from the point desired by the user: it may take time for reproduction to begin, or reproduction may start from a point different than the desired point.
To solve these problems, it is necessary to add some constraints on encoding. Adding strong constraints, however, would weaken the advantages of MPEG4-AVC, so the constraints should be minimized.
An access point AP is a position (point) accessible in random access reproduction, in which reproduction starts from an arbitrary point in the video information, e.g., a point desired by the user. For a read-only optical disc, these points are designated at the time of authoring of the disc, while for a recordable or rewritable optical disc, the points are designated by the recording device when it records video data on the disc.
In
When encoding is carried out with the above constraints, in order to start reproduction from P4, which is an access point AP, it is only necessary to read and decode the data of I1 at the start of the GOP; it is not necessary to read or decode other pictures preceding P4. Let it be assumed that the disc read-out rate is 10 Mbps, the coding rate is 10 Mbps, the coding ratio of I pictures to B pictures to P pictures is 10:6:1, and the seek time is 100 msec. In this case, the total data size of I1 and P4 will be about 1.3 Mbits under the above conditions, and the time necessary to read out all this data will be 130 msec. After reading I1, an additional 100 msec is needed to access P4, so about 230 msec will be necessary in all.
The above discussion deals with a case in which there is a single P picture access point AP in the GOP, but it is also possible to set a plurality of P pictures as access points in the GOP.
Therefore, the encoding constraints for the second and subsequent P picture access points are as follows:
When the user wants to start reproduction from P8, which is used as the second access point AP2, I1 at the start of the GOP is reproduced first. Next, P4, which is the first access point AP1, is read and decoded; then P8 is read and decoded. Subsequent reproduction can be performed in the usual manner. Differing from the case of a single access point AP, when there are a plurality of access points, it is necessary to reproduce the initial I picture of the GOP and the P pictures at access points preceding the intended access point, so the farther back in the GOP the access point is, the more time is needed before reproduction begins. Nevertheless, the pictures can be reproduced considerably more quickly than if they had been encoded without these constraints.
Under the above encoding conditions, the following index information is added to the picture data to facilitate reproduction from an access point in the interior of a GOP.
number_of_IAP in
The for loop statement following number_of_IAP in
The I_PTS_AP[IAP_id] in
The next item, I_SCN_AP, is information giving the position in the video file or on the disc of the initial I picture access point. In this embodiment, the sector offset of the I picture relative to the start of the video file is given. It will be appreciated that a byte offset or the like can be used instead of a sector offset: any information may be used that can identify the position of the I picture relative to the start of the video file, or the absolute position of the I picture on the disc. Size_of_IAP is information giving the data size of the initial I picture access point. In this example, the sector offset of the sector containing the last byte of the I picture relative to the start of the GOP including the I picture is given. From these three items of access point information, the starting point, presentation time, and data size of the initial I picture access point can be identified.
In number_of_PAP in
The next item, P_SCN_AP, is similar to the above I_SCN_AP, giving information about the position at which a P picture access point starts in a GOP, or in the video file. The sector offset relative to the start of the I picture access point is given here. The Size_of_PAP information is similar to Size_of_IAP, giving the data size of the P picture access point. The sector offset of the sector including the last byte of the P picture used as an access point, relative to the start of the P picture, is given. From the above three items of information, the video image starting position, presentation time, and data size of a P picture used as an access point can be identified.
Structuring the access point index information as described above enables positional information, temporal information, and the picture size of each access point to be identified. Next, a procedure for using the above index information to start reproduction from an access point will be described.
First, an example of the ordinary reproduction sequence will be described. When a reproduce instruction from the user is input through the user interface (I/F) 101 to the CPU 102, a command to read navigation data 1 is output to the drive controller 104, which controls the drive 103 that reads data from an optical disc. The read-out navigation data 1 are transferred via the drive controller 104 to a work memory 105. The drive controller 104 and the work memory 105 are connected to the CPU 102 via a system bus 106 comprising an address bus and a data bus, which is used to transfer commands from the CPU 102 and transfer data between blocks. The CPU 102 extracts management information regarding the program designated for reproduction by the user from the navigation data 1 deployed in the work memory 105. Based on the extracted management information, the CPU 102 instructs the drive controller 104 to read the data of the video file 3 required for reproduction from the video data 2, and the drive 103 reads out the desired data. The read-out data are temporarily stored, via the system bus 106, in a buffer memory 108 of the decoder 107, which decodes the coded data. The CPU 102 controls this process to prevent the buffer memory 108 from being exhausted or overflowing, in order to achieve reproduction without interruption in picture or sound. The data stored temporarily in the buffer memory 108 are decoded by the decoder 107 into a video signal, which is output to a display device 109 such as a TV monitor.
Next, the flow of processing in a time search, in which the user designates a time, will be described. When the user interface 101 outputs a search instruction having a time specified by the user, the CPU 102 refers to the Entry_map( ) of the navigation data 1 stored in the work memory 105. Among the I_PTS_AP in Entry_map( ), let I_PTS_AP(k) be the time information of the I picture access point closest to the time specified by the user. Among the P_PTS_AP, let P_PTS_AP(j) be the time information of the closest P picture access point (where j is an integer equal to or greater than one, indicating the j-th P picture access point counted from the I picture). The above I picture will be assumed to belong to GOP(k) 4.
The starting address of the video file 3 currently being reproduced can be recognized from the file system in the storage medium. Therefore, the absolute address of the initial I picture in the GOP(k) 4 containing the intended access point will be the starting address of the video file 3 plus the I picture position information I_SCN_AP(k). The CPU 102 instructs the drive controller 104 to read data from this absolute address. The amount of data read is equivalent to the number of sectors given in Size_of_IAP, which is the data size of the I picture. The I picture data read according to the position information and data size are temporarily loaded from the drive controller 104 into the buffer memory 108.
When the reading of the I picture is completed, the CPU 102 instructs the drive controller 104 to read data from an address determined by adding the following three addresses: the starting address of the video file, the I picture position information I_SCN_AP(k), and the position information of the next P picture access point (P_SCN_AP(1), not shown). The amount of data read is equivalent to the number of the sectors given in Size_of_PAP(1). The data of the P picture thus read according to the position information and data size is temporarily loaded from the drive controller 104 into the buffer memory 108.
In order to reproduce P picture j, it suffices to read the I picture in GOP(k) 4 and the P pictures at the following access points AP(1) to AP(j), so the above process is repeated j times. The intended access point P_PTS_AP(j) is thereby reached.
The CPU 102 calculates timings to output video from the intended access point within the shortest period, transfers data stored in the buffer memory 108 to the decoder 107, and starts decoding. Reproduction from the user-specified time is carried out in the above manner. The above description is for a time search process in which the user specifies the time at which to start reproduction, but it will be appreciated that the user can specify the picture or address at which reproduction is to start; the specified reproduction position in the present invention includes positions specified by time, address, or picture etc.
As described above, using I pictures and P pictures as access points designated as randomly accessible positions enables access points to be designated at reasonable intervals, without reducing the number of access points, even in low-bit-rate coding systems such as MPEG4-AVC.
When a picture used as an access point is a P picture, coding efficiency can be maintained by coding the picture by prediction either from a temporally preceding P picture or from the initial I picture in the GOP.
Furthermore, reproduction following the access point can proceed smoothly because pictures temporally following an I picture or a P picture used as an access point are not coded by prediction from pictures temporally preceding the I picture or the P picture used as the access point, other than the initial I picture in the GOP and other pictures used as access points.
Storing information on the optical disc or other video information recording medium giving the presentation time, position, and data size of pictures used as access points enables quick reproduction from access points in the interiors of GOPs.
According to the present embodiment, compression efficiency can be improved by increasing the GOP length without compromising random accessibility.
In the first embodiment, P_PTS_AP[IAP_id][PAP_id], which indicates relative time from the start of the GOP, was used as information giving the presentation time of the access points in the GOP, but a method of obtaining time information by using information with a smaller data size will now be described.
Display_order[IAP—id][PAP—id]=(frame_num[IAP—id][PAP—id]−frame_num[IAP—id])*M
M in this equation is a value indicating the interval between P pictures: in
P—PTS—AP[IAP—id][PAP—id]=Display_order[IAP—id][PAP—id]*3750
To make calculations such as in the two equations above possible, the following encoding constraints must be satisfied.
When the above conditions are satisfied, the presentation time of an access point can be calculated easily from the value of frame_num. The data size of frame_num is 16 bits at maximum, but it is not necessary for all the frame_num bits to be used. For example, if the maximum GOP length is five seconds, the frame frequency is 30 Hz, and M=3, then the maximum number of frames in a GOP is 150, of which 50 are P pictures. Therefore, the relative frame offset from the I picture can be calculated using the least significant six bits in frame_num.
As described above, if predetermined constraints are satisfied during encoding, the presentation time of a P picture used as an access point can be calculated easily from frame_num and the amount of data other than video data can be reduced, compared with use of PTS. Although the presentation time of an access point is calculated from frame_num in this embodiment, the presentation time of an access point can be similarly calculated from any information that indicates the decoding order.
In the first embodiment, a method of accessing an access point at a P picture in the interior of a GOP was described. In the present embodiment, a reproducing method with high image quality as well as high access speed, obtained by control of the encoding system, will be described.
In the first embodiment, the picture used for prediction of a P picture at an access point was a P picture at another access point or the initial I picture in the GOP. In general, in order to reduce the amount of code in a P picture, it should be predicted from a picture as temporally close as possible. Therefore, for the same coding bit rate, a picture can be encoded with higher image quality if it is predicted from a picture closer than the initial I picture. Access points are set at intervals such as 0.5 seconds or 1.0 second, but when a P picture is predicted from a picture this temporally remote, the amount of coded data is quite likely to increase. It can therefore be anticipated that image quality will be inferior to that obtained when no access points are set in the interiors of GOPs.
In the description of the first embodiment, bits were read from the disc at a rate comparatively close to the coding bit rate. When reading is performed at a considerably higher rate than the coding rate, the data in the interval preceding the intended access point can be read within a relatively short time without a seek operation. Therefore, if the following encoding constraints are imposed, access points can be predicted from P pictures relatively close by, and image quality will be improved:
The difference from the first embodiment is that when a P picture used as an access point is encoded it may be predicted from any P picture disposed between the I picture and the access point, provided the previous pictures necessary for decoding the P picture at the access point do not include any B pictures.
As described in the first embodiment, when no encoding constraints are imposed, eighteen pictures preceding P5 may have to be decoded in order to decode P5. The third embodiment can significantly decrease the number of pictures to be decoded.
It will assumed here that the reading rate from the disc is 10 Mbps, the coding rate is 2 Mbps, the coding ratio of I pictures to B pictures to P pictures is 10:6:1, and the time necessary for decoding a P picture is 20 msec. The amount of data from 11 to P5 is about 1.5 Mbits, which takes about 150 msec to read. Decoding up to P5 takes about 100 msec. In this case, as more time is necessary for reading than decoding, P5 can be decoded in 150 msec. In this manner, when the coding rate is considerably lower than the reading rate, access points can be reproduced quickly if the above constraints are imposed when the data are encoded. This would be impractical if the coding rate were 10 Mbps, as in the first embodiment, because then reading the data would take about 750 msec, which is five times 150 msec.
Next, a method of selecting the P pictures necessary for decoding access point P5 from the continuously read sequence of picture data will be described with reference to
As the pictures are automatically selected in the LSI chip, GOP_access_info precedes each I picture access point in the video file 3 in
In
picture_type gives attribute information that indicates if each P picture is an access point, or is necessary for decoding the next P picture access point. picture_type has the value ‘2’ when the P picture is an access point itself, the value ‘1’ when the P picture is a P picture necessary for decoding the next access point, and the value ‘0’ otherwise. From the above information, the presentation time and position of a P picture used as an access point in a GOP and the presentation time and position of the P pictures necessary for decoding the access point can be obtained.
In
In
The same effect can be obtained in this embodiment if the picture selector 110 precedes the buffer memory 108 instead of the decoder 107, and selects the picture data necessary for decoding before storing the data in the buffer memory 108.
Next, the operation in a time search in which the user specifies a presentation time will be described.
In the third embodiment, there is no need for a direct seek to a P picture access point in the GOP, because the data are read from the I picture access point to the P picture access point without a seek operation. Therefore, only information about I picture access points is necessary in Entry_map( ). In order to read an I picture at an access point, its SEI, which gives management information from the I picture access point to the next I picture access point, must be read first, so the SEI_SCN_AP information that gives the position of the SEI is recorded in Entry_map( ).
When the user specifies a time or an image to be reproduced, Entry_map( ), which is a management region separate from the content data region, is accessed, and the I picture access point closest to the specified time is identified as in the first embodiment. Next, the GOP_access_info identified by the ref_IAP_id number corresponding to this I picture access point is accessed, based on the access point information of the identified I picture, particularly on the SEI_SCN_AP information giving the position of the SEI.
A difference from the first embodiment is that data are now read continuously, starting from the address specified by I_PTS_AP(k) for the identified I picture access point. From among the data read preceding the P picture access point closest to the intended time, the P pictures necessary for decoding the access point are decoded; then the access point data and all of the data read thereafter are decoded to perform reproduction starting from the P picture access point closest to the intended time.
As described above, when the bit rate is considerably higher in reading than in encoding, access points can be specified at appropriate intervals without reducing the number of the access points, even in a low-bit-rate encoding scheme such as MPEG4-AVC, by using I pictures and P pictures as access point pictures designated as randomly accessible positions.
When a picture used as an access point is a P picture, coding efficiency and high image quality can both be achieved by coding the picture by prediction either from a temporally preceding P picture or from the initial I picture in the GOP.
Furthermore, reproduction following the access point can proceed smoothly because pictures temporally following an I picture or a P picture used as an access point are not coded by prediction from pictures temporally preceding the I picture or the P picture used as the access point, other than the initial I picture in the GOP and other pictures used as access points.
Storing information on the optical disc or other video information recording medium giving the presentation time of pictures used as access points enables quick reproduction from access points in the interiors of GOPs. Furthermore, whether a picture is necessary for decoding the next access point or not can be determined easily because attribute information indicating if the picture is necessary or not is recorded.
According to the present embodiment, compression efficiency can be improved by increasing the GOP length without compromising image quality or random accessibility.
In the case described in the third embodiment, the coding rate was lower than the reading rate, but when the coding rate does not differ significantly from the reading rate, the data reading time becomes impractically long. Suppose that the coding bit rate is 8 Mbps, and the coding ratio of I pictures to B pictures to P pictures is 4:2:1. The amount of data preceding access point P5 in
The fourth embodiment differs from the first embodiment in that the fourth constraint is added, and that in the second constraint, an access point can be predicted from a P picture not used as an access point. In the first embodiment, predictions were made from an I or P picture access point in order to avoid time-consuming seek operations. In the fourth embodiment, however, because the P pictures necessary for decoding an access point are arranged in a continuous series, predictions can be made from closer P pictures, and the amount of code at a P picture access point can be reduced.
In MPEG4-AVC, although pictures must be input to the decoder in the order in which they should be decoded, the pictures may be arranged arbitrarily when recorded. This embodiment resolves this inconsistency by proposing a new GOP structure.
In
When, as above, the pictures are not arranged in their decoding order, they must be reordered in the decoding order before being sent to the decoder.
Next, the operation in a time search in which the user specifies a presentation time will be described.
As described above, using I pictures and P pictures as access points designated as randomly accessible positions enables access points to be designated at reasonable intervals, without reducing the number of access points, even in low-bit-rate coding systems such as MPEG4-AVC.
When a picture used as an access point is a P picture, coding efficiency and high image quality can both be achieved by coding the picture by prediction either from a temporally preceding P picture or from the initial I picture in the GOP.
Furthermore, reproduction following the access point can proceed smoothly because pictures temporally following an I picture or a P picture used as an access point are not coded by prediction from pictures temporally preceding the I picture or the P picture used as the access point, other than the initial I picture in the GOP and other pictures used as access points.
When a P picture in the interior of a GOP is used as an access point, re-arranging the sequence of pictures so that the I picture and any other P pictures used for predictively encoding the P picture, that is, the I and P pictures necessary for decoding the P picture, are bunched together can shorten the access time to the access point.
Storing information on the optical disc or other video information recording medium giving the presentation time, position, and data size of pictures used as access points enables quick reproduction from access points in the interiors of GOPs.
Furthermore, re-arrangement of pictures at the time of decoding is facilitated by the recording of the decoding order of the pictures for each GOP (video unit).
According to the present embodiment, compression efficiency can be improved by increasing the GOP length without compromising image quality or random accessibility.
Increasing the GOP length has an effect not only on random access reproduction as described above; it also has considerable effect on trick reproduction modes such as fast-forward reproduction. When the GOP length is increased in the MPEG4-AVC scheme, the interval between I pictures extends to a few seconds, which adversely affects video quality in a fast-forward reproduction mode in which, for example, only I pictures are reproduced. In this embodiment, a method of fast-forward reproduction with high quality, obtained by adding improvements to the fourth embodiment, will be described. In the following explanation, descriptions of elements similar to elements in the fourth embodiment will be omitted.
In general, in order for fast-forward reproduction to proceed smoothly, not only I pictures but also P pictures in the interiors of GOPs must be reproduced. In DVD, for example, smooth fast-forward reproduction is realized by use of positional information about the first three P pictures in each GOP, which is given in the navigation data. In DVD, one GOP generally lasts 0.5 second and has about four P pictures. Therefore, if the given positional information is used to reproduce the first three P pictures in a GOP, that covers almost all of the P pictures in the GOP, making smooth fast-forward reproduction possible. When the length of the GOP is increased to a few seconds or more, however, the number of P pictures in the GOP increases, and smooth reproduction becomes difficult if reproduction is based only on the positional information about the first three pictures. Also, even if positional information for all P pictures is known, reading all the P pictures is impractical, because of the low seek speed of an optical disc. In the fourth embodiment, in order to achieve random accessibility, an access point picture is immediately followed by the P pictures necessary for decoding the next access point. The P pictures necessary for decoding access points are not exactly the same as the P pictures necessary for fast-forward reproduction, so smooth fast-forward reproduction cannot be obtained from this arrangement alone, but if the P pictures necessary for fast-forward reproduction are incorporated into in this arrangement, the P pictures in the GOP can be read without repeated seek operations.
In this embodiment, the constraints imposed on the encoding for realizing both random accessibility and fast-forward reproduction are as follows:
A method of arranging and reproducing data for high quality fast-forward reproduction, obtained by adding improvements to the fourth embodiment, will be described with reference to
The italics in
In the fourth embodiment Size_of_IAP and Size_of_PAP gave the total number of sectors necessary for random access, that is, the data size of the pictures necessary for decoding the next access point after a given access point, but in the present embodiment, Size_of_IAP and Size_of_PAP give the total data size (total number of sectors) of the pictures necessary for decoding the next access point after the current access point and the pictures necessary for fast-forward reproduction.
When a user specifies fast-forward reproduction, the transition to fast-forward reproduction occurs at the start of the next GOP. The operation through the reading of the amount of data specified by Size_of_IAP is exactly the same as in random access. The pictures thus read are all decoded, in order to decode from the next access point, but the pictures actually reproduced are only those with attribute flags specifying fast-forward reproduction. When the reading of the amount of data specified by Size_of_IAP is completed, a seek operation is performed to read the amount of data specified by Size_of_PAP from the next P picture access point. Similar operations are repeated until an instruction to stop fast-forward reproduction is received from the user. Smooth fast-forward reproduction is realized by the above operations. The operations during random access are exactly the same except that, based on the attribute values, the pictures necessary for decoding the next access point are selected from among the pictures read according to the data size specified by Size_of_IAP or Size_of_PAP; a description will be omitted. Incidentally, although the preceding description assumes that the decoder 107 in
A method of improving image quality in fast-forward reproduction by adding further improvements to the fourth embodiment has been described in the present embodiment, but fast-forward reproduction can also be realized in the third embodiment, in which the pictures do not have to be reordered, by adding an attribute flag indicating whether or not a picture is used in random access and/or fast-forward reproduction, in place of the picture_type item shown in the third embodiment.
As described above, using I pictures and P pictures as access points designated as randomly accessible positions and as pictures necessary for trick reproduction such as fast-forward reproduction enables access points to be designated at reasonable intervals and makes smooth trick reproduction possible, even in low-bit-rate coding systems such as MPEG4-AVC.
When a picture used as an access point or a picture necessary for trick reproduction is a P picture, coding efficiency and high image quality can both be achieved by coding the picture by prediction either from a temporally preceding P picture or from the initial I picture in the GOP.
Furthermore, reproduction following the access point can proceed smoothly because pictures temporally following an I picture or a P picture used as an access point are not coded by prediction from pictures temporally preceding the I picture or the P picture used as the access point, other than the initial I picture in the GOP and other pictures used as access points.
When a P picture in the interior of a GOP is used as an access point or is necessary for trick reproduction, or both, re-arranging the sequence of pictures so that the I picture and any other P pictures used for predictively encoding the P picture, that is, the I and P pictures necessary for decoding the P picture, are bunched together can shorten the access time to the access point.
Storing information on the optical disc or other video information recording medium giving the presentation time, position, and data size of pictures used as access points enables quick reproduction from access points in the interiors of GOPs.
Recording attribute information indicating whether a picture is a picture necessary for decoding of the next access point, a picture necessary for trick reproduction such as fast-forward reproduction, or a picture necessary for both of these purposes permits easy determination of whether a picture is a P picture used in random access or trick reproduction, and facilitates reordering when pictures are reordered during decoding.
As shown in the fourth embodiment, reordering of pictures at the time of decoding is facilitated by the recording of the decoding order of the pictures for each GOP (video unit).
According to the present embodiment, compression efficiency can be improved by increasing the GOP length without compromising image quality, random accessibility, or smooth trick reproduction.
Number | Date | Country | Kind |
---|---|---|---|
2004-195476 | Jul 2004 | JP | national |
2004-205250 | Jul 2004 | JP | national |
2004-214080 | Jul 2004 | JP | national |
2004-229683 | Aug 2004 | JP | national |
2004-238482 | Aug 2004 | JP | national |
This application is a Continuation of application Ser. No. 10/569,603, filed on Feb. 24, 2006, now U.S. Pat. No. 7,706,668 and for which priority is claimed under 35 U.S.C. §120. Application Ser. No. 10/569,603 is the national phase of PCT International Application No. PCT/JP05/11342 filed on Jun. 21, 2005 under 35 U.S.C. §371. This application claims priority of Application Nos. 2004-195476, 2004-205250, 2004-214080, 2004-229683, and 2004-238482 all filed in Japan on Jul. 1, 2004, Jul. 12, 2004, Jul. 22, 2004, Aug. 5, 2004, and Aug. 18, 2004, respectively, under 35 U.S.C. §119; the entire contents of all are hereby incorporated by reference.
Number | Name | Date | Kind |
---|---|---|---|
5377051 | Lane et al. | Dec 1994 | A |
20020041758 | Seki et al. | Apr 2002 | A1 |
20050008328 | Kawa et al. | Jan 2005 | A1 |
Number | Date | Country |
---|---|---|
1388862 | Feb 2004 | EP |
8-45249 | Feb 1996 | JP |
08-273304 | Oct 1996 | JP |
09-046712 | Feb 1997 | JP |
09-098430 | Apr 1997 | JP |
2001-285877 | Oct 2001 | JP |
2003-319340 | Nov 2003 | JP |
2005-012390 | Jan 2005 | JP |
WO-9730544 | Aug 1997 | WO |
WO-0225949 | Mar 2002 | WO |
Number | Date | Country | |
---|---|---|---|
20080063360 A1 | Mar 2008 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 10569603 | US | |
Child | 11978570 | US |