Buffer overflow and underflow control for facsimile transmission

Abstract

In a facsimile system, buffer underflow is prevented by forcing the transmit pointer to wait at a current end-of-line (EOL) symbol and transmit fill bits until there are at least two EOL symbols stored in the buffer. Buffer overflow is controlled by selectively dropping lines if the write pointer approaches too close to the transmit pointer and the difference between successive lines is sufficiently small (selective line-dropping), dropping lines whenever a permitted maximum buffer size is exceeded, and gradually changing the permitted distance between transmit and write pointers and permitted maximum buffer size as the image is scanned to accomplish a smooth buffer flushing.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The invention relates generally to facsimiles, and, more particularly, to controlling buffer overflow and underflow for facsimile transmission over mobile communication channels by using fill bits, selective line-dropping, drastic line-dropping, and smooth buffer-flushing.

2. Description of the Related Art

It is known that facsimile transmission time can be reduced by using compression techniques before transmission of facsimile documents scanned into a sending facsimile machine. The compressed facsimile documents are stored in a buffer before being transmitted via a communication network to a receiving facsimile machine. Image compression techniques are particularly important in facsimile transmission in mobile communications because of the relatively high cost for air time. Further, in mobile communication networks it is known to use a circular buffer to store the digitized and compressed facsimile document before transmission. However, the circular buffer can experience buffer overflow or underflow problems which degrade the accuracy of the facsimile document as reproduced by the receiving facsimile machine.

To transmit standard Group 3 facsimile (fax) documents over mobile communications networks in a cost-efficient manner, it is necessary to employ secondary image compression techniques to reduce the amount of information necessary to describe the image. When secondary image compression is employed for a standard G3 fax machine, it is essential to implement a method of buffer control to control the flow of data bits between the compression algorithm and the mobile channel. Controlling the flow of data is required since image compression gains can vary significantly from image scan line to scan line which will not necessarily be coincident to the transmission channel transmission rate. For instance, if there is a high degree of correlation between a current scan line and its previous scan line, then the compression gain for the current scan line will be much higher than if there were a low degree of similarity between adjacent scan lines. As a result, statistical variations in these compression gains lead to fluctuations in the rate at which encoded bits of information become available for transmission. Since the transmission channel requires a steady flow of bits at a constant rate, a buffering mechanism is required to interface the compression algorithm and the communications channel. However, the buffer can suffer from buffer underflow and buffer overflow which creates illegible and/or incomplete documents to be transmitted to the receiving fax machine.

Buffer underflow results when, over a period of time, the compression process generates data bits at a slower rate than can be transmitted over the channel. Buffer underflow must be suitably controlled to maintain proper operation and transmission of the fax document or the fax document will not be legible as printed by the receiving tax machine. Buffer overflow results when, over a period of time, the compression algorithm generates bits at a higher rate than can be transmitted over the channel. If not controlled, buffer overflow can lead to a loss of data and a disruption in the transmission of the image, each contributing to an illegible and/or incomplete document as printed by the receiving fax machine. Buffer overflow is controlled by generating, just before the buffer capacity is exceeded, a control signal which causes the facsimile system to delete the preceding scan line.

Further, traditional facsimile compression techniques suffer from a problem of clipping the bottom of highly detailed documents because there is a maximum period of 1.6 seconds available for emptying the buffer at the end of each transmitted document page. Any untransmitted data that remains in the buffer after the 1.6 seconds available for emptying the buffer is deleted from the buffer and not transmitted, allowing the buffer to begin loading data for the next transmission page. The more highly detailed the transmission document, the more data that will be deleted from the end of the transmission document page, and the more clipping of the printed document that occurs at the receiving facsimile machine. This problem occurs in traditional facsimile compression techniques when sending highly detailed documents regardless of the facsimile machine's ability to control the buffer to prevent underflow or overflow during transmission.

SUMMARY OF THE INVENTION

In view of the above, one object of the present invention is to provide an improved facsimile system.

Another object of the invention is to provide a facsimile system which enables the use of secondary compression techniques for efficient transmission of facsimile documents over mobile networks without significant buffer underflow or buffer overflow.

Another object of the present invention is to provide a facsimile system which can transmit complex documents without losing data as a result of the bottom of a page being clipped.

Another object of the present invention is to provide a facsimile system which can transmit complex documents with a minimal amount of information loss confined to only extremely detailed documents, while still maintaining high overall image intelligibility.

Another object of the present invention is to provide a process for facsimile transmission which improves facsimile document transmission accuracy by controlling a buffer mechanism to prevent data underflow and/or overflow.

According to one aspect of the present invention, buffer underflow is prevented by forcing the buffer's transmit pointer to wait at the previous end-of-line (EOL) symbol, e.g., a bit sequence consisting of 11 0's followed by a 1 (i.e., 000000000001), until at least one subsequent EOL symbol is present in the buffer. While the transmit pointer waits at the most recent EOL symbol, the channel is then stuffed with fill bits, bits with value 0. This technique is simple to implement, causes no loss in the performance of the compression algorithm, and cures the problem of underflow which causes loss of synchronization and distortion of the facsimile document.

According to another aspect of the present invention, buffer overflow and page cutoff is prevented using the buffer control mechanisms of selective line-dropping, drastic line-dropping, and smooth buffer-flushing processes to control the flow of data information in the transmission of facsimile documents across mobile networks. The present invention uses selective line-dropping to solve the buffer overflow problem and drastic line-dropping and smooth buffer-flushing processes to solve the page cutoff problem.

For selective line-dropping, a 20,000 bit circular buffer is provided with various zones to indicate distinct regions of bit locations within which the buffer's write pointer can be relative to the buffer's transmit pointer in the circular buffer. The transmit pointer reads bits but of the circular buffer to be sent across the channel; the write pointer writes bits into the circular buffer after the compression algorithm. In a preferred embodiment, there are five distinct zones designated in the circular buffer which include the safety zone, low critical zone, middle critical zone, high critical zone, and cutoff zone, each zone respectively corresponding to the write pointer getting closer to the transmit pointer and thereby increasing the probability of buffer overflow (i.e., write pointer overcoming the transmit pointer). If the write pointer is located anywhere within the safety zone, no overflow control is necessary. If the write pointer falls within any one of the three critical zones (low, middle or high), buffer overflow algorithms are initiated. First, the current scan line and previous scan line are compared with one another to determine their bit difference value, and this value is compared with a threshold associated with the particular critical zone, with the critical zones having higher thresholds as the transmit pointer is approached more closely by the write pointer. If the difference value between the two scan lines does not exceed the threshold, for that critical zone, then the current scan line and previous scan lines are considered similar enough for line-dropping to take place. Otherwise, no line-dropping occurs and the present scan line is simply encoded.

If a scan line is determined to be dropped, a special line-dropping bit is set to 1 and sent to the decoder. At the receiving facsimile system, the line-dropping mode is detected by monitoring the code bit and the previously decoded scan line is duplicated at the current scan line's position in the output document.

When a highly detailed document is scanned, compression gains are extremely low, the probability of buffer overflow is extremely high, and the write pointer may actually fall within the cutoff zone. Should this situation arise, an end of page (EOP) sequence, consisting of 6 EOL sequences, is immediately transmitted to the receiving facsimile machine. This signals the end of document transmission and the termination of any further decoding. The remainder of the document, from the point of cutoff, is then replaced with blank lines, ensuring that the resulting document at the receiving facsimile machine is not filled with garbage, but rather, implicating to the person receiving the fax, that an error has occurred during the transmission process.

To eliminate the page cutoff problem, smooth buffer-flushing is initiated at the start of document processing, where after a certain percentage of the incoming image has been scanned, the maximum number of bits allowed to exist in the buffer is reduced. This is performed in a stair-step manner such that the maximum number of bits allowed in the buffer is reduced as the percentage of the present page scanned increases.

Within each percentage region of a page, selective line-dropping is customized for that specific maximum buffer capacity. Hence, as the permitted maximum buffer capacity gradually decreases over the course of a document, the sizes of the safety zone and three critical zones increase, thereby increasing the probability of selective line-dropping and reducing the probability of buffer overflow. Within each of the percentage regions, drastic line-dropping is initiated if the number of bits between the write pointer and transmit pointer exceeds the maximum number of bits allowed to exist in the buffer for that region. Drastic line-dropping is the same as selective line-dropping except that all scan line bits written into the circular buffer are consecutively dropped until the distance between the write and transmit pointers is less than the maximum buffer threshold.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and other objects, features, and advantages of the present invention will become more apparent from the following description of one preferred embodiment thereof taken in conjunction with the accompanying drawings, wherein:

FIG. 1 is a diagram showing the land earth facsimile station in accordance with the present invention.

FIG. 2 illustrates a CCITT highly detailed document used to test the performance of the present invention.

FIG. 3 illustrates a CCITT highly detailed document which has been clipped at the end of the page due to the use of traditional facsimile compression techniques on a standard G3 facsimile machine.

FIG. 4 is a diagram illustrating the data format for one scan line in accordance with the present invention.

FIG. 5 is a flow chart illustrating the control buffer underflow process in accordance with the present invention.

FIG. 6 is a diagram showing the various zones ot the circular buffer in accordance with the present invention.

FIG. 7 is a flow chart illustrating the control buffer overflow process in accordance with the present invention.

FIG. 8 illustrates a CCITT highly detailed document simulated with the rough buffer-flushing process of the present invention.

FIG. 9 illustrates a CCITT highly detailed document simulated with the smooth buffer-flushing process of the present invention.

FIG. 10A is an original test document as scanned into the facsimile machine, used to illustrate the improved facsimile document accuracy and legibility of the present invention.

FIG. 10B is the processed test document of FIG. 10 A and illustrates the simulation of the buffer control processes of the present invention.

FIG. 11A is an original test document as scanned into the facsimile machine, used to illustrate the improved facsimile document accuracy and legibility of the present invention.

FIG. 11B is the processed test document of FIG. 11 A and illustrates the simulation of the buffer control processes of the present invention.

FIG. 12A is an original test document as scanned into the facsimile machine, used to illustrate the improved facsimile document accuracy and legibility of the present invention.

FIG. 12B is the processed test document of FIG. 12 A and illustrates the simulation of the buffer control processes of the present invention.

FIG. 13A is an original test document as scanned into the facsimile machine, used to illustrate the improved facsimile document accuracy and legibility of the present invention.

FIG. 13B is the processed test document of FIG. 13 A and illustrates the simulation of the buffer control processes of the present invention.

DESCRIPTION OF THE PREFERRED EMBODIMENT

Referring to FIG. 1 , the present invention comprises a standard G3 fax machine 1 and a secondary image compression technique in which the fax image is first decoded by the T.4 Huffman algorithm 2 and then compressed using a secondary compression algorithm 3. The T.4 Huffman algorithm is that shown in CCITT Recommendation T.4 , “Standardization of Group 3 Facsimile Apparatus for Document Transmission”, Red Book, Vol. VII, Facsimile VII.3, Malaga-Torremolinos 1984, pp. 16-3 1.

A 1×2 Maximum Differences compression algorithm 3 is used for the compression process. The 1×2 Maximum Difference compression algorithm of the present invention is that shown in “ Compression of Facsimile Graphics for Transmission over Digital Mobile Satellite Circuits”, Spiros Dimolitsas and Frank corcoran, IEEE Military Communications Conference, MILCOM 91, Washington D.C., Nov. 4-8, 1991. However, the present invention is equally applicable with other image compression algorithms that are capable of compressing binary images.

After secondary image compression 3 has been completed, the resulting bits of information are sent to an output buffer 4 where they await channeling to the satellite receiver 5 . Such buffers are typically implemented as DSP circular buffers so that their finite storage space may be continually reused. This finite storage capacity also represents a limit on the maximum number of bits that can be stored in the buffer, before buffer overflow occurs.

The rate at which bits of information are written into the circular buffer, via the write pointer (wp), is dependent upon the amount of information stored in the incoming scan line since different scan lines contain different amounts of information. As a result, bits are written into the buffer as bursts of information rather than at a constant rate. However, the rate at which these bits of information are read from the buffer (for transmission over the satellite channel), via the transmit pointer (tp), is constant. For example, the transmission rate is 3,200 bps for the Inmarsat-M satellite system.

For highly detailed documents, where compression gains are significantly lower compared to less detailed documents, the rate at which the write pointer writes into the circular buffer 4 is often greater than the rate at which the transmit pointer reads. Thus, the problem of buffer overflow occurs when the write pointer has the opportunity to overtake the position of the transmit pointer before the transmit pointer has had a chance to read what was written into the buffer 4 . The results are often catastrophic as the facsimile document at the receiving Mobile Earth Station (in the case of the Inmarsat-M system) becomes heavily distorted, or completely untransmitted from the point of overflow. The selective line-dropping process of the present invention was developed to address this problem and incorporates a special code bit, known as the line-dropping bit. This process triggers various line-dropping routines to increase the relative distance between the two pointers, thereby keeping the write pointer at a “safe” distance behind the transmit pointer and eliminating any possibility of buffer overflow.

In developing a complete solution to buffer overflow the additional problem of page clipping is also addressed. Due to reception of the T.30 end-of-page (EOP) confirmation message of a document transmission, there exists a maximum period of 1.6 seconds available for emptying the buffer 4 , i.e., buffer-flushing, at the end of each page for the document being transmitted. Since the transmit pointer reads at a constant rate of 3,200 bps for the facsimile machine of the preferred embodiment, it can be calculated that no more than 5,000 bits of information (approximately 3,200 bps×1.6 seconds) can remain in the buffer 4 and be accurately transmitted at the end of document transmission. If there are more than 5,000 bits of information, buffer-flushing cannot be completed and the resulting document will become “clipped” or devoid of information towards the bottom of the page equivalent to the number of leftover bits in the buffer.

For highly detailed documents such as CCITT test document 7 illustrated in FIG. 2 , this is a severe problem since approximately 30,000 bits remain in the buffer 4 at the end of document transmission if no methods of buffer control are exercised. When a maximum of 5000 bits in the buffer is finally flushed after document transmission, this results in the loss of about 25,000 bits of information and approximately 15% of the document is clipped as illustrated in FIG. 3 . The present invention overcomes this clipping problem by using drastic line-dropping and smooth buffer-flushing algorithms to gradually reduce the number of bits in the buffer to a maximum of 5,000 bits. This ensures that the buffer can be fully flushed at the end of document transmission with only a minimal loss of information for the overall document.

The processes of the present invention have been rigorously tested on over seventy-five facsimile documents, including the 8 standard CCITT test documents, to yield optimal performance. Although employing the buffer control techniques as shown in the present invention will result in some overall document degradation, the amount of distortion is limited only to highly detailed documents and result in minimal loss of intelligibility. By employing such processes, the immediate advantages include the following: 1) feasibility of employing secondary compression techniques for efficient transmission of fax over mobile networks, 2) transmission of complex documents without bottom-of-page clipping, and 3) minimal amount of information loss confined to only extremely detailed documents while still maintaining high overall image intelligibility. Finally, as will be clear to one skilled in the art, the present invention is flexible and applicable to a wide variety of other systems such as Inmarsat Mini-M, North American digital cellular systems, European GSM, and AFRTS.

As previously mentioned, the standard G3 fax image is decoded into image scan lines using T.4 Huffman decoding 2 , which are, in turn, compressed by the secondary compression technique such as the 1×2 Maximum Differences compression algorithm 3. The resulting bits of information are then stored in a circular buffer 4 , awaiting transmission across the satellite channel 5 . The boundary between successive scan lines is marked by an end-of-line (EOL) symbol, which is a binary sequence of 11 zeros and a single one (000000000001). The EOL symbol is immediately followed by two special code bits. The first bit is a mode bit which indicates to the decoder whether the current scan line was encoded in a 1-dimensional mode or a 2-dimensional mode. The second bit is a line dropping bit which indicates if the current scan line bits were “dropped” or left untransmitted to the decoder. The capability of dropping a scan line and indicating this action to the decoder is a key component of the buffer overflow control mechanism. A detailed description of the buffer underflow control and buffer overflow control processes of the present invention follows.

The buffer underflow control process will now be described with reference to FIGS. 4 and 5 . Buffer underflow is prevented by ensuring that the write pointer (wp) is always ahead of the transmit pointer (tp). At the beginning of the current page to be processed, the write and transmit pointers are each at the 0 th memory location in the circular buffer. Each scan line is then encoded as shown in FIG. 4 . As each scan line is compressed, the corresponding bits of information are written into the buffer, thereby advancing the write pointer. FIG. 5 illustrates the process steps of decoding and compression as a single step 10 . Normally, the transmit pointer also advances at a constant rate (3,200 bps for Inmarsat-M).

FIG. 5 illustrates the additional steps in the buffer underflow control process which are as follows. First, in step 11 the present scan line data is written into the buffer 4 and the write pointer advances. When the transmit pointer reaches an EOL symbol, step 12 is entered and the transmit pointer is only allowed to advance further if there is at least one more EOL symbol in the buffer between the transmit and the write pointers. If this condition is not satisfied, the transmit pointer is forced to wait at the current EOL symbol until the next EOL is written into the buffer after repeating steps 0 and 11 . During the waiting period when the transmit pointer is fixed at the current EOL, at step 13 , the transmission channel is stuffed with fill bits which are bits of binary value 0. The decoder understands that any sequence of zero valued bits arriving after complete decoding of a scan line and just preceding an EOL symbol are fill bits, and can consequently discard these bits. On the other hand, if at step 12 there are 2 or more EOL symbols detected, the transmit pointer is allowed to proceed normally by transmitting previously decoded and compressed line data while the next line is scanned, decoded, and compressed as shown in step 14 .

Note that an advantage of sending fill bits before the EOL symbol is that these bits can be easily recognized and are stripped at the decoder. On the other hand, if fill bits were instead inserted at arbitrary positions within the scan line, rather than at EOL symbols, it would become necessary to dedicate bit patterns for the transmission of the “fill” information. Since such patterns cannot be employed in the encoding operation, this would significantly reduce the efficiency of the encoding algorithm.

The buffer overflow control process will now be described with reference to FIGS. 6 and 7 . The buffer overflow mechanism of the present invention consists of a 20,000 bit circular buffer 4 within which various zones are set for which the write pointer 21 can come to within the transmit pointer 22 . According to FIG. 6 , there are five distinct regions in the circular buffer. The safety zone limits the write pointer 21 to within anywhere between 20,000 bits to 3,000 bits counter-clockwise, behind the transmit pointer 22 . If the write pointer 21 is located anywhere within this region, it is considered a “safe” data bit distance behind the transmit pointer 22 such that no overflow control is necessary. The reason this region is considered safe is since every scan line contains a maximum number of 1728 pels (picture elements), when the next scan line is encoded and compressed, there is no possible way for the write pointer 21 to overtake the transmit pointer 22 if there is at least a 3,000 bit difference between the two pointers.

If the write pointer 21 is between 2,999 bits to 2,000 bits behind the transmit pointer, the write pointer 21 is said to reside in the low critical zone; between 1,999 bits to 1,000 behind the transmit pointer 22 represents the middle critical zone; between 999 to 85 bits behind the transmit pointer 22 represents the high critical zone, and between 84 bits and 0 bits behind the transmit pointer 22 represents the cutoff zone.

Referring to FIG. 7 , the process steps for the selective line dropping algorithm include steps 62 - 73 and will hereinafter be described. If the write pointer 21 falls within any one of the three critical zones, buffer overflow algorithms are initiated by step 64 for the low critical zone, step 65 for the middle critical zone, and step 66 for the high critical zone. Then, in step 69 , the current scan line and previous scan line are compared with one another to determine the pel (picture element) percentage difference, or amount of non-correlation, between the two scan lines. The pel percentage difference is defined as, $$ pel ⁢   ⁢ percentage ⁢   ⁢ difference = # ⁢   ⁢ of ⁢   ⁢ pels ⁢   ⁢ different ⁢   ⁢ between ⁢   ⁢ previous ⁢   ⁢ and ⁢   ⁢ current ⁢   ⁢ scan ⁢   ⁢ line total ⁢   ⁢ # ⁢   ⁢ of ⁢   ⁢ pels

Next, in step 70 , the pel percentage difference is compared to the pel percentage threshold specified for that critical zone.

In the low critical zone, the pel threshold is set at, for example, 15%; in the middle critical zone, the pel percentage threshold is set at, for example 20%, and in the high critical zone the pel percentage threshold is set at, for example, 25%. If the pel percentage difference between the two scan lines is less than or equal to the pel percentage threshold specified for that critical zone, then the current scan line and previous scan lines are considered similar enough for line-dropping to take place as shown in steps 71 and 72 . For instance, assume that after the k th scan line has been encoded and compressed, the write pointer 21 and transmit pointer 22 adjust themselves so that the final position of the write pointer 21 is 1,500 bits counter-clockwise behind the transmit pointer 22 , in the middle critical zone. If the pel percentage difference between the k th scan line and the (k−1) th scan line is less than or equal to 20%, line-dropping takes place in steps 71 and 72 , followed by encoding of the (k+1) th scan line in step 73 . Otherwise, no line-dropping occurs; both the k th and (k+1) th scan lines are simply encoded.

When line-dropping occurs in one of the three critical zones, the transmit pointer 22 advances normally at 3,200 bps while the current position of the write pointer 21 is brought back to its previous position in the buffer as shown in step 71 . As a result, the information bits of the current scan line are not transmitted to the decoder. Although this strategy results in a small loss of information equivalent to the number of bits in the dropped scan line, the advantages from line-dropping are an increase in the distance between the two pointers and thus, the elimination of any buffer overflow.

After a scan line has been dropped, a specific 14-bit sequence consisting of a 12-bit end-of-line (EOL) sequence, 000000000001, followed by the mode bit, and followed by the line-dropping bit set to 1, are sent to the decoder as indicated by step 72 . The line-dropping bit is a special code bit reserved for indicating line-dropping mode to the decoder. The bit is normally reset to 0 and transmitted immediately after. the EOL sequence as part of each encoded line. At the receiving facsimile system, the line-dropping mode is detected by monitoring the code bit. When the decoder of the receiving facsimile system reads the line-dropping bit as 1, line-dropping has occurred and the previously decoded scan line is duplicated at the current scan line's position in the output document. If the line-dropping bit is set to 0 (default), signifying that no line-dropping has occurred, the current scan line is simply reconstructed in the resulting image.

In some highly detailed documents where compression gains are extremely low, the write pointer 21 may actually fall within the cutoff zone as shown in step 67 . Should this situation arise, an EOP sequence (6 EOL sequences) is immediately transmitted across the satellite receiver 5 as shown in step 68 . This also signals the end of document transmission and the termination of any further decoding. The remainder of the document, from the point of cutoff, is then inserted with blank lines, ensuring that the resulting document at the receiving facsimile machine is not filled with garbage, but rather, implicating to the person receiving the fax transmission that an error has occurred during the transmission process.

It should be noted that such an event seldom, if ever, occurs since the current Huffman encoding 2 and Maximum Differences compression techniques 3 yield sufficient enough compression gains which when combined with the buffer control algorithms, very rarely allow for the write pointer 21 to enter the cutoff zone. For the seventy-five sample documents tested with the present invention, including the extremely detailed CCITT test document 7 , none of the images were clipped due to cutoff.

Selectively dropping a scan line within each specific critical zone has two major advantages. First, although line-dropping and duplicating the previously decoded line result in a small loss of information, the positive trade-off is that only a slight amount of distortion is generated as opposed to clipping the entire document from the point of overflow if no forms of overflow control were exercised. The second advantage of selective line-dropping is that it allows for more evenly distributed distortions across the entire image instead of highly clustered distortions in one central region. This increases the document's overall intelligibility and attractiveness.

FIG. 7 also illustrates the drastic line-dropping process and smooth buffer-flushing process as shown in steps 51 - 61 and hereinafter described. The drastic line-dropping algorithm is used in conjunction with the smooth buffer-flushing algorithm to reduce the number of bits in the buffer to 5,000 bits at the end of document transmission. This ensures that complete buffer-flushing can occur within the timing constraint of 1.6 seconds between pages for the facsimile system of the preferred embodiment. Smooth buffer-flushing at step 53 is initiated at the start of document processing, where after a certain percentage X of the incoming page image has been scanned, the maximum number of bits allowed to exist in the buffer 4 is reduced. Table 1 below displays the maximum buffer sizes, Y, allowed for various ranges of scanned document, identified as percentage X. Note that a test document is representative of one page of a standard facsimile document.

TABLE 1
Percentage X Maximum
of           Number of
Document     Bits Y in
Scanned      Buffer
0-5          20,000 bits
5-10         17,000 bits
10-15        15,000 bits
15-25        13,000 bits
25-35        11,500 bits
35-45        10,000 bits
45-55        8,500 bits
55-65        7,500 bits
65-75        6,500 bits
75-85        6,000 bits
85-95        5,500 bits
95-100       5,000 bits

Within each percentage X region of the document, the selective line-dropping algorithm is customized for that specific maximum buffer capacity Y to achieve smooth buffer-flushing. Hence, as the buffer capacity gradually decreases, for example, from 20,000 bits to 5,000 bits over the course of a document (page), the relative sizes of the safety zone and three critical zones increase, thereby increasing the probability of selective line-dropping and reducing the probability of buffer overflow. The process steps for smooth buffer-flushing and drastic line-dropping are as follows.

The percentage X region location of the document scan is determined at step 53 and provided for adjusting the three critical zone bases. If no percentage of the document page has previously been stored in the buffer 4 the process will proceed to step 59 and determine the new transmit pointer 22 and write pointer 21 positions. Alternatively, if a portion of the document page has been previously stored in the buffer 4 , the process proceeds to step 54 . As shown in step 54 , within each of the percentage X of document scanned regions, drastic line-dropping is initiated if the number of bits between the write pointer 21 and transmit pointer 22 exceeds the maximum number of bits Y allowed to exist in the buffer 4 for that region. If the number of bits between the write pointer 21 and the transmit pointer 22 does not exceed the maximum number of bits Y allowed in the buffer 4 for that region, the process proceeds to process the next scan line in step 58 . Drastic line-dropping as defined herein is analogous to selective line-dropping except all scan line bits written into the circular buffer are consecutively dropped until the bit difference between the write pointer 21 and transmit pointer 22 is less than the maximum buffer threshold as indicated at step 54 .

For instance, assume that after the k th scan line has been processed, 88% of the document has been scanned and that the number of bits between the two pointers is 7,500 bits. Since the number of bits in the buffer 4 , 7,500 bits, exceeds the current maximum buffer threshold of 5,500 bits in the 88% region, drastic line-dropping becomes necessary. Upon encoding the (k+1) th scan line, the transmit pointer 22 will advance normally at 3,200 bps, the write pointer 21 will be brought back to its previous position in the buffer at step 56 , and the special 14-bit sequence consisting of an EOL, mode bit, and line-dropping bit having the line-dropping bit set to 1 at step 57 , will be transmitted to the channel decoder. This process continues until the number of bits between the two pointers falls below 5,500 bits.

It is possible that in highly detailed documents with low compression gains, the result of many consecutive drastic line-droppings will be a region of high distortion consisting of a cluster of duplicated scan lines. This is an inevitable consequence for extremely detailed documents since the current scan line must be continually dropped and the previous scan line reconstructed in order to reduce the number of bits in the buffer to the required threshold. However, for approximately 90% of the documents tested, there were no occurrences of selective or drastic line-droppings. As a result, these algorithms displayed near transparent image quality.

One of the major advantages of smooth buffer-flushing and drastic line-dropping is the gradual reduction of maximum buffer capacities over time as opposed to sharp decreases. As can be understood, employing a 20,000 bit buffer for the first 80% of a document and suddenly switching to a 5,000 bit buffer could generate tremendous distortion since many scan lines would need to be drastically dropped in order to reduce the buffer size from 20,000 bits to 5,000 bits at the 80% threshold. The result of such “rough buffer-flushing” is a centralized region of severe distortion. FIG. 8 illustrates CCITT test document 7 simulated with the rough buffer-flushing conditions described above. It can be seen at the 80% region that there exists a clearly visible section of high distortion due to many consecutive line-droppings.

On the other hand, if CCITT test document 7 were processed with the smooth buffer-flushing algorithm of the present invention, there would be fewer drastic line-droppings since switching between adjacent buffers would involve fewer bits to be left untransmitted. As a result of gradually or smoothly flushing the buffer, fewer clustered line-droppings result, thereby giving the document an overall higher degree of attractiveness and legibility as illustrated by CCITT test document 7 simulated with the smooth buffer-flushing algorithm in FIG. 9 . The resulting image is much more pleasing to the human eye since there does not exist one centralized region of severe distortion, but rather, all visible degradations have been successfully scattered throughout the entire document. Likewise, this process corrects the clipping problem (compare FIG. 3 with FIG. 7 ).

3. Sample Document Tests

Table 2 below displays the results of simulating the 8 CCITT test documents, and 3 other highly detailed documents through the buffer underflow and buffer overflow algorithms. Note that CCITT test documents 4 and 7 are of highest detail since their compression gains are the lowest.

TABLE 2
		Maximum			Number of	Number of
	Gain	Buffer Size	Bits Left in	Bits Saved	Drastic	Total
	(including	(w/o line-	Buffer at	by Line-	Line-	Line-
Document	fill bits)	dropping)	EOP	dropping	Droppings	Droppings
CCITT-1	1.50	1143	0	0	0	0
CCITT-2	1.50	367	0	0	0	0
CCITT-3	1.50	1019	0	0	0	0
CCITT-4	1.41	22507	1729	17600	42	57
( FIGS. 10A &
10B)
CCITT-5	1.50	893	0	0	0	0
CCITT-6	1.50	548	0	0	0	0
CCITT-7	1.36	33351	2174	29674	40	90
CCITT-8	1.50	400	0	0	0	0
vbtypes	1.42	20739	1791	15939	24	34
( FIGS. 11A &
11B)
txtt10	1.42	16948	857	11982	21	27
( FIGS. 12A &
12B)
patd	1.44	8059	4290	3135	13	13
( FIGS. 13A &
13B)

From the data in Table 2, it can be observed that for documents whose compression gains are high (i.e., 1.50), buffer overflow is not a problem, as the number of drastic line-droppings and total line-droppings is inconsequential. In such documents, buffer underflow is the major dilemma and the problem is prevented with the insertion of fill bits as previously described for the buffer underflow control process. However, as more detail is introduced to documents, thereby shrinking the amount of redundancy between scan lines and decreasing the overall compression gain, the number of dropped scan lines increases. Because highly detailed documents are packed with so much information per scan line, the number of bits in the buffer can grow arbitrarily large during the encoding and compression processes.

In an effort to reduce this difference to under 5,000 bits as to meet the 1.6 seconds EOP timing constraint, the physical evidence of drastic line-dropping can be observed in test documents CCITT 4 , CCITT 7 , vbtypes, txtt 10 , and patd, all of which experienced selective and drastic line-dropping and contain dispersed regions of extremely small distortion. FIGS. 10 through 13 illustrate these documents simulated with the buffer control processes of the preferred embodiment. Figures labeled with the A designation represent the original document as scanned into the facsimile machine and figures labeled with the B designation represent test documents as processed by the aforementioned buffer underflow control process and buffer overflow control process of the preferred embodiment. It can be clearly seen that the overall quality of the resulting images in FIGS. 10B , 11 B, 12 B, and 13 B is very high. Although the simulated documents in these figures are not exact duplicates of the originals, they are highly legible relative to the facsimile documents transmitted by a standard G3 facsimile machine and significantly reduce the amount of total transmission time, hence cost, due to the ability to effectively use secondary image compression techniques. The occurrence of selective and drastic line-dropping will be inevitable for extremely detailed documents. However, the algorithms of the present invention are considered crucial and practical to control what might otherwise result in inefficient means of secondary image compression, transmission delays, transmission errors, bottom-of-page document clipping, and distortion, if no methods of buffer control are provided.

As can be appreciated by one skilled in the art, the invention is relevant to facsimile systems in general, and more specifically to a number of mobile communications networks, such as mobile satellite (Inmarsat-M, Inmarsat Mini-M) systems and digital cellular (North American Digital Cellular) systems.

While the invention has been described in conjunction with the preferred embodiment, it will be understood that it is not intended to limit the invention to this particular embodiment. To the contrary, the invention is intended to cover alternatives, modifications, and equivalents, which may be included within the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method of controlling transmission in a facsimile system for transmitting data in the form of a plurality of lines of an image, each line of data ending with a known code, said method comprising the steps of:storing said data in a buffer prior to transmission; and restricting output of data from said buffer until at least two of said known code are present in said buffer.

2. A method according to claim 1, further comprising the step of stuffing a transmission channel with fill bits while output from said buffer is being prevented.

3. A method according to claim 2, wherein said output buffer is a circular buffer having a transmit pointer and a write pointer and the step of restricting output of data comprises restricting advancement of said transmit pointer while said write pointer advances to write a present facsimile data line into said circular buffer.

4. A method according to claim 2, wherein said fill bits have a predetermined data value.

5. A method of controlling transmission in a facsimile system for transmitting data in the form of a plurality of lines of an image, said system including a buffer for storing data to be transmitted, said data being written to said buffer at locations corresponding to a write pointer and read out of said buffer from locations corresponding to a transmit pointer, said method comprising the steps of:defining a threshold corresponding to a degree of difference between a current line and a previous line; changing said threshold during transmission of a given image; comparing said current and previous lines; and dropping said current line in accordance with the results of said comparing step.

6. A method according to claim 5, wherein said dropping step comprises dropping said current line if said threshold is not exceeded.

7. A method according to claim 5, wherein said defining step comprises defining a plurality of zones each corresponding to a respective range of distances between said write pointer and transmit pointer, and defining a different threshold for each said zone.

8. A method of controlling transmission in a facsimile system for transmitting data in the form of a plurality of lines of an image, said system including a buffer for storing data to be transmitted, said data being written to said buffer at locations corresponding to a write pointer and read out of said buffer from locations corresponding to a transmit pointer, said method comprising the steps of:defining a plurality of zones each corresponding to a respective range of distances between said write pointer and transmit pointer; defining for each zone a respective threshold corresponding to a degree of difference between a current line and a previous line; comparing said current and previous lines; and dropping said current line in accordance with the results of said comparing step and in accordance with the distance between said write and transmit pointers.

9. A method according to claim 8, further comprising the step of changing the size of at least one of said zones as a function of the amount of image transmitted.

10. A method according to claim 8, further comprising the step of transmitting a line drop indicator each time that a line is dropped.

11. A method according to claim 8, wherein each page of facsimile image ends with an end of page marker, said method further comprising the step of terminating further transmission of an image after an end of page marker is detected if a threshold is exceeded which corresponds to a closest range of distances between said write and transmit pointers.

12. A method according to claim 8, wherein said plurality of zones comprises four zones.

13. A method of controlling transmission in a facsimile system for transmitting data in the form of a plurality of lines of an image, said system including a buffer for storing data to be transmitted, said data being written to said buffer at locations corresponding to a write pointer and read out of said buffer from locations corresponding to a transmit pointer, said method comprising the steps of:defining a maximum permitted amount of data in said buffer; adjusting said maximum in accordance with an amount of image transmitted; and dropping a line of said image if said maximum is exceeded.

14. A method according to claim 13, wherein said defining step comprises defining a plurality of zones each corresponding to a respective percentage completion of said image, and defining a different maximum for each said zone.

15. A method according to claim 14, wherein a zone corresponding to a largest amount of completion of said image has the lowest maximum.

16. A facsimile system for transmitting data in the form of a plurality of lines of an image, each line of data ending with a known code, comprising:a buffer for storing said data prior to transmission; and a transmission control mechanism restricting output of data from said buffer until at least two of said known codes are present in said buffer.

17. A system according to claim 16, further comprising a stuffing mechanism for stuffing a transmission channel with fill bits while output from said buffer is being prevented.

18. A system according to claim 17, wherein said output buffer is a circular buffer having a transmit pointer and a write pointer and said transmission control mechanism restricts output of data by restricting advancement of said transmit pointer while said write pointer advances to write a present facsimile data line into said circular buffer.

19. A facsimile system for transmitting data in the form of a plurality of lines of an image, said system including a buffer for storing data to be transmitted, said data being written to said buffer at locations corresponding to a write pointer and read out of said buffer from locations corresponding to a transmit pointer, said system further comprising:means for defining a threshold corresponding to a degree of difference between a current line and a previous line; means for changing said threshold during transmission of a given image; means for comparing said current and previous lines; and means for dropping said current line in accordance with the results of said comparing step.

20. A system according to claim 19, wherein said means for defining defines a plurality of zones each corresponding to a respective range of distances between said write pointer and transmit pointer, and said thresholds being different for each said zone.

21. A system according to claim 19, wherein said means for dropping drops said current line if said threshold is not exceeded.

22. A system according to claim 19, wherein said means for changing changes said threshold as a function of the amount of said image transmitted.

23. A facsimile system for transmitting data in the form of a plurality of lines of an image, said system including a buffer for storing data to be transmitted, said data being written to said buffer at locations corresponding to a write pointer and read out of said buffer from locations corresponding to a transmit pointer, said system further comprising a transmit control mechanism which compares said current and previous lines and drops said current line in accordance with the results of said comparison and in accordance with a distance between said write and transmit pointers.

24. A system according to claim 23, wherein said transmit control mechanism defines a plurality of zones each corresponding to a respective range of distances between said write pointer and transmit pointer, with each zone having a respective threshold corresponding to a degree of difference between a current line and a previous line, said threshold mechanism dropping a line if said threshold is not exceeded in any zone.

25. A system according to claim 23, further comprising means for transmitting a line drop indicator each time that a line is dropped.

26. A system according to claim 23, wherein each page of facsimile image ends with an end of page marker, said transmit control mechanism terminating further transmission of an image after an end of page marker is detected if a threshold is exceeded which corresponds to a closest range of distances between said write and transmit pointers.