Method of processing video image data for use in the storage or transmission of moving digital images

• US 5,067,015 A
• Filed: 06/18/1990
• Issued: 11/19/1991
• Est. Priority Date: 06/28/1989
• Status: Expired due to Fees

• Current Assignee: British Aerospace, Inc.
• Original Assignee: British Aerospace, Inc.
• Inventors: Combridge, Brian L., Timson, Daniel R. E., Robson, David
• Primary Examiner(s): Britton, Howard W.
• Application Number:
  US07/539,814
• Time in Patent Office: 519 Days
• Field of Search: 358/133, 358/136
• US Class Current: 375/240.11
• CPC Class Codes: H04N 19/10 using adaptive coding
• CPC Class Codes: H04N 19/115 Selection of the code volume for a coding unit prior to coding
• CPC Class Codes: H04N 19/146 Data rate or code amount at the encoder output
• CPC Class Codes: H04N 19/186 the unit being a colour or a chrominance component
• CPC Class Codes: H04N 19/53 Multi-resolution motion estimation; Hierarchical motion estimation
• CPC Class Codes: H04N 19/61 in combination with predictive coding
• CPC Class Codes: H04N 19/63 using sub-band based transform, e.g. wavelets
• CPC Class Codes: H04N 19/635 characterised by filter definition or implementation details
• CPC Class Codes: H04N 19/645 by grouping of coefficients into blocks after the transform

Abstract

A method of processing video image data for use in the storage or transmission of moving digital video images, includes decomposition of each frame image into an equivalent set of sub-images containing information on different length scales, derivation of encoded prediction rules from the decomposed data to approximate each sub-image using data from the immediately preceding compressed sub-image and data from the current sub-image to produce compressed data of an image for subsequent storage or transmission to a receiver and reconstruction, after retrieval or receipt, by comparison of the compressed data received for the current frame image with the reconstruction of the preceding frame image.

Claims

1. A method of processing video image data for use in the storage or transmission of moving digital video images, comprising the steps of:

decomposing each frame image into an equivalent set of sub-images containing information on different length scales,

deriving encoded prediction rules from the sub-images decomposed in said decomposing step;

approximating each sub-image using data from the immediately preceding compressed sub-image and data from the current sub-image to produce compressed data of an image for subsequent storage or transmission to a receiver; and

reconstructing de-compressed current frame data, after retrieval or receipt, by comparison of the compressed data received for the current frame image with a reconstruction of the preceding frame image.

2. A method of processing video image data for use in the storage or transmission of moving digital video images, including the steps of examining a current frame image, sub-band decomposing the current frame image by bandpass filtering to produce a set of sub-images, dividing each sub-image in the current frame image set into a plurality of blocks, comparing the blocks for each sub-image in the current frame image set with the blocks in the approximation of the corresponding sub-image in the preceding frame image set for similarities and variations in illumination intensity, generating for each block of each sub-image of the current frame image set a prediction rule and errors resulting therefrom to produce for each sub-image of the current frame image partially compressed data which is the sum of the predicted sub-image and corresponding error image, and further compressing the partially compressed data for each sub-image of the current frame image by encoding to produce fully compressed current frame image data, for subsequent passage or transmission to storage or a receiver and reconstruction of the retrieved or received fully compressed current frame image data.

3. A method according to claim 2, in which sub-band decomposition is an iterative operation carried out more than once, with each successive bandpass sub-image being generated at each iteration and containing information from the upper half of the remaining spatial frequency spectrum in each dimension.

4. A method according to claim 3, in which the set of bandpass sub-images is completed by a final lowpass sub-image containing information from the whole of the remaining spatial frequency spectra.

5. A method according to claim 4, in which the first bandpass sub-image has the same number of pixels as the current frame image, with each successive sub-image being sub-sampled by a factor of two in each dimension relative to the previous sub-image.

6. A method according to claim 5, in which before subsequent processing the sub-band decomposition is exactly invertible so that the current frame image is recoverable without error from the set of sub-images.

7. A method according to claim 6, including the step of quantizing the pixel values of the sub-images.

8. A method according to claim 7, wherein the prediction rule for each block of each sub-image of the current frame image involves `stationary match` in which a block is sufficiently similar to the spatially coincident block in the approximation of the corresponding preceding frame sub-image or `displaced match` in which a block is sufficiently similar to a block displaced from the spatially coincident block in the approximation of the corresponding preceding frame sub-image or `mean value` in which the pixel values of a block are estimated by the average of those values or `intraframe` in which each pixel value in a block is estimated on the basis of the approximations of neighbouring pixel values in the current frame sub-image.

9. A method according to claim 8, in which the prediction rule for each block of each sub-image of the current frame image includes predictions for the pixel values of the block in the form of the spatially coincident values from the approximation of the corresponding preceding frame sub-image or the appropriately displaced values or the average of the pixel values in the current frame sub-image block or the interpolations of the approximations for neighbouring pixel values in the current frame sub-image.

10. A method according to claim 9, in which the errors for each block of each sub-image of the current frame image are the differences between the pixel values of the actual and predicted current frame sub-image block.

11. A method according to claim 10, in which before encoding the errors are processed by quantisation, deletion of spatially isolated non-zero values and/or random deletion of non-zero values, further to increase the data compression effect resulting from subsequent encoding.

12. A method according to claim 11, in which prior to passage or transmission of fully compressed current frame image data to storage or to a receiver, the total number of bits of fully compressed data is counted, and if the total number falls outside a predetermined range, the total number of bits of the fully compressed data for the next current frame image is varied accordingly.

13. A method according to claim 12, including the steps of passage or transmission of fully compressed current frame image data to storage or a receiver, and reconstruction of the retrieved or received fully compressed current frame image data, in which reconstruction the prediction rules and processed error images are decoded, the approximations to each of the sub-images are recovered and the sub-band decomposition of the original current frame image is inverted.

14. A method according to claim 13, in which the approximations to each of the sub-images are recovered by applying the prediction rules to the corresponding approximations recovered for the immediately preceding frame image and adding the respective error image pixel values thereto.

15. A method according to claim 13, in which the sub-band decomposition is inverted by increasing the number of pixels in the recovered lowpass sub-image by a factor of two in each dimension, adding the pixel values of the smallest recovered bandpass sub-image to produce an approximation to the lowpass image for that level, and repeating said increasing and adding steps for each consecutive level in the current frame image sub-band decomposition until an approximation of the original current frame image is produced.