Process and device for the compression of image data by mathematical transformation effected at low cost, particularly for the transmission at a reduced rate of sequences of images

• US 5,031,038 A
• Filed: 04/18/1990
• Issued: 07/09/1991
• Est. Priority Date: 04/18/1989
• Status: Expired due to Fees

• Current Assignee: Etat Francais, Telediffusion De France SA
• Original Assignee: Etat Francais, Telediffusion De France SA
• Inventors: Guillemot, Christine, Duhamel, Pierre, Carlac'h, Jean-Claude
• Primary Examiner(s): Britton, Howard W.
• Application Number:
  US07/510,752
• Time in Patent Office: 447 Days
• Field of Search: 358/133, 358/136, 364/725
• US Class Current: 375/240.2
• CPC Class Codes: G06F 17/147 Discrete orthonormal transforms, e.g. discrete cosine transform, discrete sine transform, and variations therefrom, e.g. modified discrete cosine transform, integer transforms approximating the discrete cosine transform G06F17/145 takes precedence
• CPC Class Codes: H04N 19/60 using transform coding

Abstract

Process and device for the compression of image data by a mathematical transformation implemented at low cost, particularly for the transmission at reduced speed of sequences of images.

The objective is to provide an economical method in "silicon area - processing speed" cost to implement a mathematical transformation of the DCT type, on arrays of pixels, suitable for image compression.

According to the invention, the method consists essentially of combining a permutation (1) of each block of pixels with a series of one dimensional FFT (211, 213). A series of variants is obtained by adding or not adding steps of matrix calculation (22) of reconstitution of a complex 2D-FET, of rotations (3) of reconstitution of the DCT, and/or of suppression of the imaginary coefficients.

Claims

1. Process of compression of data representative of an image signal, of the type consisting of decomposing the image into blocks forming arrays of pixels, of effecting a mathematical operation of transformation on each array of pixels according to a transform making it possible to obtain a resulting array of elements representative of the activity of the image, and of eliminating in each resulting array the elements of little psychovisual relevance by thresholding in relation to an array of reference thresholds, process characterized in that the said mathematical operation of transformation consists of the steps of:

transposing the elements of the array of pixels by permutation of the indices of the said elements, so as to symmetrize the signal;

applying an interlacing of two series of one dimensional real value FFT, bearing successively on each line of the permuted array of pixels, then on each of the columns of the intermediate array resulting from the first series of one dimensional FFT.

2. Process according to claim 1 characterized in that at least part of the imaginary coefficients that are obtained after at least one of the two series of one dimensional real value FFT are cancelled.

3. Process according to claim 1 wherein said interlacing of two series of one dimensional real value FFT are replaced with a two dimensional real or complex value FFT.

4. Process according to the claim 1 characterized in that at least one of the two one dimensional FFT is followed by an operation of matrix transposition with bit reverse.

5. Process according to claim 4 characterized in that on the transformed array resulting from the two interlaced one dimensional FFT a matrix calculation is effected designed to reconstitute the array transformed by complex value FFT corresponding to the said initial transposed array of pixels.

6. Process according to the claim 5 characterized in that the said array transformed by complex value FFT is subjected to a treatment of rotations, so that a resulting array is obtained corresponding to the real value DCT transform of the initial array of pixels.

7. Process according to claim 5 wherein said interlacing of two series of one dimensional real value FFT are replaced with a two dimensional real or complex value FFT.

8. Process according to any of the claims 1 to 6 characterized in that the law of permutation of the indices of the elements of the array of pixels is combined with a scanning of the array being chosen in the group comprising a line to line scan of the array and a zigzag scan of the array.

9. Process according to claim 3 or 7 characterized in that at least part of the imaginary coefficients that are obtained after said two-dimensional real or complex value FFT are cancelled.

10. Device to implement the process according to any one of the claims 1 to 6, 3, 7 or 2 characterized in that it comprises two distinct memory units, the said memory units cooperating with a data path effecting in pipeline the interlaced operations of reading coefficients in the memories, the calculation operations on the said coefficients, and the interlaced writing operations of the resulting coefficients in the memory units, so as to effect the steps of the process by successive passages of the coefficients in the said data path.