Color image data compression and recovery apparatus based on neural networks

• US 5,041,916 A
• Filed: 02/06/1990
• Issued: 08/20/1991
• Est. Priority Date: 02/07/1989
• Status: Expired due to Term

• Current Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Yoshida, Kunio, Tanaka, Takehisa, Saitoh, Mie, Naka, Motohiko
• Primary Examiner(s): Coles, Sr., Edward L.
• Assistant Examiner(s): Grant, II, Jerome
• Application Number:
  US07/474,396
• Time in Patent Office: 560 Days
• Field of Search: 358/429, 358/432, 358/261.1, 358/261.2, 358/261.3, 358/261.4, 358/262.1, 358/80, 358/433
• US Class Current: 382/233
• CPC Class Codes: G06T 9/002 using neural networks
• CPC Class Codes: H04N 1/648 Transmitting or storing the primary (additive or subtractive) colour signals; Compression thereof H04N1/642 - H04N1/646 take precedence
• CPC Class Codes: H04N 11/042 Codec means

Abstract

A data compression and recovery apparatus compresses picture element data of a color image by expressing two primary color values of each picture element as a set of parameter values of a neural network in conjunction with reference color data values of a corresponding block of picture elements. Date recovery is achieved by inputting each block of reference color values to a neural network while establishing the corresponding set of parameter values in the network, to thereby obtain the original pair of encoded primary color values for each of successive picture elements. The third primary color can be used as the reference color.

Claims

1. A data compression and recovery apparatus for compressing and subsequently recovering data expressing a source image as an array of tricolor data sets each consisting of three primary color data values of a corresponding picture elment of said source image, the apparatus comprising a data compression apparatus and a data recovery apparatus, in which said data compression apparatus comprises:

a first neural network means for adaptive processing;

first block extraction means for extracting successive blocks of m by n tricolor data sets from said source image data array, where m and n are fixed integers, said blocks respectively mutually overlap by one picture element position, and each of said reference color data values is derived from a corresponding one of said tricolor data sets, and for applying the reference color data values of an extracted block as parallel inputs of said first neural network means, while outputting a pair of primary color data values of an object picture element at a fixed position within said extracted block;

comparator means for comparing said pair of primary color data values with output data values produced from said first neural network means;

learning control means for receiving a signal representing respective current states of a set of network parameter values of said first neural network means and receiving comparison signals produced from said comparator means, and for repetitively generating and establishing in said first neural network means updated sets of network parameter values, until a predetermined degree of convergence is indicated by said comparator means output signals; and

output control means for transmitting as compressed output data the reference color data values of all of said source image together with, for each of said blocks, a corresponding set of network parameter values which provide said degree of convergence;

and in which said data recovery apparatus comprises;

a second neural network means for adaptive processing;

input control means for receiving said transmitted reference color data values and sets of network parameter values;

second block extraction means for extracting successive blocks of m by n reference color data values from the received reference color data values, and;

means for establishing successive ones of said sets of network parameter values in said second neural network means, in synchronism with applying corresponding ones of said blocks of reference color data values as inputs to said second neural network means, to obtain as output data values from said second neural network means successive pairs of primary color data values for respective picture elements of said source image.

3. A data compression and recovery apparatus according to claim 1, in which prior to transmission of said reference color data values and network parameter values, the reference color data values and network parameter values outputted from said data compression apparatus are supplied to a local data recovery apparatus and recovered data from the local data recovery apparatus are displayed by a display device, and in which a requisite number of repetitions of said procedure is determined by visual comparison of said source image with an output image produced by said display device.

4. A data compression and recovery apparatus according to claim 1 or 2, in which each of said reference color data values is a fixed one of said primary color data values of a picture element.

5. A data compression and recovery apparatus according to claim 1 or 2, in which each of said reference color data values is a function of a plurality of said primary color data values of a picture element.

6. A data compression and recovery apparatus according to claim 1 or 2, in which each of said reference color data values is a combination of a fixed one of said primary color data values of a picture element with a value that is a function of a plurality of the primary color data values of the picture element.

2. A data compression and recovery apparatus for compressing and subsequently recovering data expressing a source image as an array of tricolor data sets each comprising three primary color data values of a corresponding picture element of said source image, the apparatus comprising a data compression apparatus and a data recovery apparatus, in which said data compression apparatus comprises:

a first neural network means for adaptive processing;

first block extraction means for extracting successive blocks of m by n reference color data values from said source image data array, where m and n are fixed integers, said blocks successively mutually overlap by one picture element position, and each of said reference color data values is derived from a corresponding one of said tricolor data sets, and for supplying the data values of an extracted block as parallel inputs to said first neural network means while outputting a pair of primary color data values of an object picture element which is at a fixed position within said extracted block;

comparator means for comparing said pair of primary color data values with output data values produced from said first neural network means;

learning control means coupled to receive a signal representing respective current states of a set of network parameter values of said first neural network means and to receive comparison output signals produced from said comparator means, for repetitively executing a procedure whereby data values of an extracted block are supplied to input terminals of said first neural network means, an updated set of network parameter values is derived based on a current set of network parameter values and on said comparator output signals, the updated set of network parameter values are established in the first neural network means, and data values of a succeeding block are applied as inputs to the first neural network means, said procedure being repeated for all of said source image a plurality of times, until a desired degree of convergence is obtained between said neural network output data values and pairs of object picture element primary color data values for substantially all of said blocks; and

output control means for transmitting, as compressed output data, the reference color data values of all of said source image together with a final set of network parameter values established at completion of said procedure repetitions;

and in which said data recovery apparatus comprises;

a second neural network means for adaptive processing of identical configuration to said first neural network means;

input control means for receiving said transmitted receiving color data values and set of network parameter values;

second block extraction means for extracting successive blocks of m by n reference color data values from the received reference color data values, and;

control means for fixedly establishing said set of network parameter values in said second neural network means and for successively applying said blocks of reference color data values as inputs to said second neural network means, for thereby obtaining as output data values from said second neural network means successive pairs of primary color data values for respective picture elements of said source image.

7. A data compression and recovery apparatus for compressing and subsequently recovering data expressing a source image as an array of tricolor data sets each comprising three primary color data values of a corresponding picture element of said source image, the apparatus comprising a data compression apparatus and a data recovery apparatus, in which said data compression apparatus comprises:

a first neural network means for adaptive processing;

means for successively deriving, for each picture element of said source image, a corresponding set of parameter values for said first neural network means such that values respectively substantially identical to two of the primary color data values of said each picture element are generated from said first neural network means when reference color data values of a block of picture elements including said each picture element are inputted to said first neural network means; and

means for transmitting said sets of parameter values and respective reference color data values of all of said source image;

and in which said data recovery apparatus receives aid transmitted parameter values and reference color data values and comprises;

a second neural network means for adaptive processing;

means for successively establishing said sets of parameter values in said second neural network means in synchronism with inputting to said second neural network means the reference color data values of successive blocks respectively corresponding to said sets of parameter values, for thereby successively obtaining as output values from said second neural network means said first and second primary color values of all of said picture elements.

9. A data compression and recovery apparatus according to claim 7 or 8, in which each of said reference color data values is a fixed one of said primary color data values of a picture element.

10. A data compression and recovery apparatus according to claim 7 or 8, in which each of said reference color data values is a function of a plurality of said primary color data values of a picture element.

11. A data compression and recovery apparatus according to claim 7 or 8, in which each of said reference color data values is a combination of a fixed one of said primary color data values of a picture element with a value that is a function of a plurality of the primary color data values of the picture element.

8. A data compression and recovery apparatus for compressing and subsequently recovering data expressing a source image as an array of tricolor data sets each comprising three primary color data values of a corresponding picture element of said source image, the apparatus comprising a data compression apparatus and a data recovery apparatus, in which said data compression apparatus comprises:

a first neural network means for adaptive processing;

means for successively deriving for all of the picture elements of said source image a single set of parameter values for said first neural network means such that, for each of said picture elements, values respectively substantially identical to two of the primary color data values of said each picture element are generated from said first neural network means when reference color data values of a block of picture elements including said each picture element are inputted to said first neural network means; and

means for transmitting said set of parameter values and respective reference color data values of all of said source image;

and in which said data recovery apparatus receives said transmitted parameter values and reference color data values and comprises;

a second neural network means for adaptive processing;

means for establishing said set of parameter values in said second neural network means, and for successively inputting to said second neural network means reference color data values of successive ones of said blocks, for thereby successively obtaining as output value from said second neural network means said first and second primary color values of all of said picture elements.