×

Learning Filters For Enhancing The Quality Of Block Coded Still And Video Images

  • US 20090034622A1
  • Filed: 08/01/2008
  • Published: 02/05/2009
  • Est. Priority Date: 08/01/2007
  • Status: Active Grant
First Claim
Patent Images

1. A method for block-wise encoding still or video images comprising the steps of:

  • a) encoding input image data composed of pixels having original pixel values using a block-wise encoder to obtain encoded image data;

    b) decoding at least a subset of the encoded image data to obtain a decoded image composed of pixels having decoded image values;

    c) identifying in the decoded image a plurality of coding block boundaries;

    d) forming training data comprising a plurality of cross-boundary filter support blocks (CBFSBs), each CBFSB comprising a group of N>

    1 adjacent pixels of the decoded image including adjacent pixels on both sides of a block boundary;

    e) generating a set of classification rules based on the training data for classifying each of the CBFSBs into at least one of a plurality of classes, so that each class is adaptively defined by a respective set of the classification rules;

    f) for each class defined in step (e), generating a set of filter coefficients associated with the class and defining a class-optimized filter that provides a prediction value of a selected CBFSB pixel in dependence upon decoded values of CBFSB pixels based on an original value of the selected CBFSB pixel for each CBFSB in said class; and

    ,g) providing the sets of filter coefficients for each class and the classification rules as an output for subsequent adaptive class-based filtering of the decoded image data for suppressing blocking artefacts.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×