Method and apparatus for adaptive facsimile compression using a two dimensional maximum likelihood predictor

• US 4,325,085 A
• Filed: 06/09/1980
• Issued: 04/13/1982
• Est. Priority Date: 06/09/1980
• Status: Expired due to Term

• Current Assignee: Hughes Electronics Corporation (Rtx Corporation)
• Original Assignee: Digital Communications Corp.
• Inventors: Gooch, Robert P.
• Primary Examiner(s): BRITTON, HOWARD W
• Application Number:
  US06/157,609
• Time in Patent Office: 673 Days
• Field of Search: 358/260, 358/261, 358/280, 358/133, 340/347 DD, 364/515
• US Class Current: 382/234
• CPC Class Codes: G06T 9/004 Predictors, e.g. intraframe, interframe coding
• CPC Class Codes: G06T 9/005 Statistical coding, e.g. Huffman, run length coding
• CPC Class Codes: H04N 1/417 using predictive or differential encoding

Abstract

Data compression, for either a storage or transmission, of facsimile information is effected employing a two dimensional, non-contiguous prediction matrix. A data stream representing the data to be compressed is input to a buffer which is of a size to store sufficient information for prediction purposes. A predictor is responsive to the buffer for producing a predicted data representation from a plurality of data units comprising a two dimensional matrix. A selector is responsive to the data unit employed in the prediction process for making a select/non-select determination. For those data units which are selected, a comparator compares the predicted status of the data unit with the actual status of the data unit. At least one run length encoder is responsive to the comparator for run length encoding successive correct predictions and a following incorrect prediction. An output buffer is provided for storing the run length encoded output of the run length encoder as well as representations of the non-selected data units. By using plural run length encoders, each can be optimized for the encoded data by correlating prediction difficulty with code length, i.e., easy predictions are encoded by long code words and hard predictions are encoded with short words. Further, the unselected class data units correspond to most difficult predictions and these are not coded.

Claims

1. A data compression device for information scanned in two dimension and represented by a serial stream of pixel representations, comprising:

first buffer means for storing at least some of said pixel representations sufficient to represent a portion of said information extending in both said dimensions;

predictor means coupled to said first buffer means for generating, from at least three pixel representations, a predicted pixel representation for a further pixel, said pixels forming a pixel set extending in both said dimensions,

selection means responsive to said at least three pixel representations for selecting, or not selecting, said further pixel for compression encoding,

comparing means for comparing said predicted representation of a selected pixel with an actual representation of said pixel from said first buffer means, to determine if said prediction is or is not correct,

run length encoding means responsive to said comparing means for encoding successive correct predictions and a following incorrect prediction as at least one multibit word, and second buffer means responsive to said run length encoding means and said predictor means for storing said at least one multibit word and at least one unencoded pixel representation corresponding to an unselected pixel.

2. The device of claim 1 wherein said predictor means predicts a pixel representation from only eight prior pixels.

3. The device of claim 1 or 2 in which said predictor means operates on seven pixels to predict a representation of a ninth pixel forming a set of nine pixels extending through at least three parallel pixel lines with at least two pixels in one of said lines separated by a pixel which is not in said set of nine pixels.

4. The device of claim 1 which includes pre-processor means coupled to an input of said first buffer means for deleting pixel runs of length one.

5. The device of claim 1 or 4 in which said predictor means comprises

a read only memory addressed by selected outputs of said first buffer means with a data entry for a set of addresses defined by possible outputs of said first buffer means.

6. The device of claim 5 in which said read only memory stores, for each unselected pixel, a fixed binary quantity.

7. The device of claim 5 in which said read only memory stores, for each unselected pixel, a representation of a statistically most likely representation for said further pixel.

8. The device of claim 5 in which said read only memory stores, at each addressed location, data corresponding to one of plural control signals.

9. The device of claim 5 in which said run length encoding means comprises

first and second run length encoders, each with a data and control input, said data input at each said first and second encoders coupled to said comparing means and each said control input coupled to a different one of said plural control signals, said first run length encoder capable of higher data compression than said second run length encoder.

10. The device of claim 1 in which said serial stream is comprised of plural lines of pixels, each line consisting of lines from plural images separated by an image marker,

and in which said second buffer means includes means to signal an impending overflow,

control means responsive to said impending overflow condition to terminate output from said first buffer means at a determined location in said serial stream, corresponding to said image marker.

11. The device of claim 1 which includes

control means responsive to said serial data stream to generate format signals, said control means including counting means, counting a predetermined number of pixels for generating a line reset signal,

said first buffer means comprising delay means including means responsive to said line reset signal to clear said delay means.

12. The device of claim 1 wherein said first buffer means comprises

first and second delay means, with an output of said first delay means coupled to an input of said second delay means, and wherein said second delay means includes means responsive to said line reset signal to clear said second delay means.

13. The device of claim 11 or 12 in which said counting means produces said line reset signal after counting a number of pixels representing a number of lines in the range of 20-40 lines.

14. The device of claim 1 which further includes

transmitter means responsive to compressed data from said second buffer means for transmitting said compressed data,

receiving means for receiving said compressed data,

third buffer means responsive to said receiving means for writing said compressed data and for, on command, reading said compressed data,

decoding means coupled to said third buffer means responsive to said compressed data for decoding said compressed data and for, at times, supplying said command to said third buffer means,

predictor means and fourth buffer means, said fourth buffer means storing expanded data and providing selected pixels to said predictor means, said predictor means responsive to said decoding means and said fourth buffer means to generate expanded data and coupled to an input of said fourth buffer means,

counting means coupled to said fourth buffer means for counting pixel representations in a pixel line, and

control means responsive to an erroneous number of pixels in a line for deleting said line and replacing said line with a copy of another line.

15. The apparatus of claim 14 in which said

fourth buffer means comprises third and fourth delay means each with an output and input, with an output of said third delay means coupled to an input of said fourth delay means and in which said control means comprises a bistable device, normally in one condition and responsive to a control signal from said counting means counting an excessive number of pixels, for switching to a second condition, and switching means, normally coupling said third delay means output to said fourth delay means input and responsive to said bistable device in said second condition to couple said fourth delay means output to said fourth delay means input.

16. A method of compressing a serial digital data stream representing pixels of an image scanned in two dimensions at a resolution sufficient to negate the importance of single pixel runs comprising the steps of:

eliminating single pixel runs, from said serial data stream,

selecting a plurality of pixels from a set of pixels representing an image portion extending in two dimensions from which to predict the condition of a last remaining pixel in said set,

predicting, from said plurality of pixels, the condition of said remaining pixel and determining whether or not run length encoding is appropriate for said predicted pixel,

comparing said predicted pixel with said actual pixel and run length encoding correct/incorrect predictions,

filling a buffer with a sequence of run length codes representing run length encoding of said correct/incorrect predictions and following said codes with one or more signals representing pixels for which run length encoding was determined to be inappropriate.

17. The method of claim 16 in which said selecting step includes selecting a plurality of only eight pixels with which to predict the state of a ninth pixel in a nine pixel set.

18. The method of claim 17 in which said nine pixel set extends over portions of at least three lines and in which at least one said line includes two pixels from said set separated by a pixel not included in said set of nine pixels.

19. The method of claim 16 in which said predicting step includes the steps of:

determining, for each different plurality of eight pixels, the most likely state of a ninth pixel in a set of nine pixels and the entropy of said ninth pixel,

comparing said entropy with a first reference and, if said first reference exceeds said entropy, determining that run length encoding is appropriate or otherwise determining that run length encoding is inappropriate.

20. The method of claim 19 in which said predicting step further includes the steps of:

comparing said entropy, if less than said first reference, with a second reference, and determining a first run length encoding or second run length encoding, is appropriate for said pixel if said entropy is less or greater than said second reference, respectively.

21. The method of claim 20 which includes the further step of recording, at a different location, an entry corresponding to each different plurality of pixels, and indication of the most likely state for said remaining pixel, and in which said step of predicting includes reading said recorded entry to determine the most likely condition for said remaining pixel.

22. The method of claim 21 in which said recording step includes recording, along with each entry, an indication of results of said comparisons with said first and second references.

23. A method of transmitting image representing data in compressed form from a transmitter to a receiver comprising the steps of:

raster scanning a field of view in two dimensions and digitizing samples obtained by said scanning to produce a serial stream of digital pixel representations,

arranging two independent images in said field aligned so that each line of said raster covers corresponding portions of both said independent images,

compression encoding at least some information from said serial stream, to produce a compression encoded stream,

transmitting said compression encoded stream to a receiver,

compression decoding said compression encoded stream at a receiver, and

recreating, at said receiver, at least one of said two independent images.

24. The method of claim 23 which includes the further step of:

buffering said compression encoded stream prior to said transmission, said buffering comprising the steps of;

writing said compression encoded stream into at least two buffers in an order of receipt as dictated by said raster,

reading from both said buffers signals representing a first image prior to reading signals from said buffers representing a second image.

25. The method of claim 24 which includes the further steps of:

additionally buffering signals read from said two buffers, in a third buffer,

monitoring the contents of said third buffer, and

aborting transfer of signals representing said second image in the event said third buffer contents approach capacity of said third buffer.