Method and device for spatial filtering of digital images decoded by transformation of block

• US 5,422,964 A
• Filed: 04/09/1993
• Issued: 06/06/1995
• Est. Priority Date: 04/17/1992
• Status: Expired due to Term

• Current Assignee: Sagem Communications (Safran SA)
• Original Assignee: SAT Corporation (Kratos Defense & Security Solutions Incorporated)
• Inventors: Devimeux, Daniel, Jolivet, Jean-Claude
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): ANDERSON, DAVID
• Application Number:
  US08/045,666
• Time in Patent Office: 788 Days
• Field of Search: 382/54, 382/56, 382/1, 382/6, 348/470, 348/576, 348/627, 348/671, 348/674, 348/688, 358/433, 358/455, 358/463, 358/461
• US Class Current: 382/226
• CPC Class Codes: H04N 19/527 Global motion vector estimation
• CPC Class Codes: H04N 19/86 involving reduction of coding artifacts, e.g. of blockiness

Abstract

Coding and decoding of digital images by transformation of pixel blocks reduce the data bit rate. Transformation of blocks leads to discontinuities in brightness level for pixels bordering on block boundaries. These discontinuities are detected for each given bordering pixel by a change of sign of two level gradients respectively computed with the levels of two pixels adjacent to the bordering pixel. A brightness level of the bordering pixel is then corrected by a discontinuity correction factor evaluated from a discontinuity divergence equal to the absolute value of the difference between the brightness level of the bordering pixel and an arithmetical average of brightness levels of the two adjacent pixels, as a function of a quantization pitch associated with a block to which belongs the bordering pixel. The block is then corrected in its entirety to attenuate the differences of average brightness levels between two adjacent blocks generated by the coding and decoding.

Claims

1. A method for filtering a digital decoded image obtained after encoding and decoding by transformation and quantization of pixel blocks of an initial image of a dimension N×

N, N being an integer, said digital decoded image being formed into a plurality of blocks each comprised of N parallel segments of pixels having respective brightness levels and being separated, in line and column, from adjacent blocks by respective boundaries, said method comprising the steps for each half-segment of said N segments in a given block separated from a corresponding collinear half-segment in one of said adjacent blocks by a respective one of said boundaries of;

(a) detecting a discontinuity in the brightness level with respect to a pixel of said each half-segment bordering on said respective boundary when two discrete gradients of brightness level depending respectively on brightness levels of two pixels adjacent to said bordering pixel have opposite signs;

(b) evaluating in response to said brightness level discontinuity a discontinuity divergence equal to an absolute value of a difference between the brightness level of said bordering pixel and an arithmetical average of brightness levels of said two adjacent pixels;

(c) correcting said brightness level of said bordering pixel into a corrected brightness level respectively;

(1) by subtracting from said brightness level of said bordering pixel a discontinuity correction factor (FCD) which depends on said discontinuity divergence and on a quantization pitch (QZ) used in said encoding and decoding by transformation of said given block, when said brightness level of said bordering pixel is more than said arithmetical average, and

2) by adding said discontinuity correction factor to said brightness level of said bordering pixel when said brightness level of the bordering pixel is less than said arithmetical average, and;

(d) reconstituting a digital reconstituted image which is substantially analogous to said initial digital image and which includes said pixel blocks of said digital decoded image with said bordering pixels having brightness levels uncorrected and corrected, respectively, according to said detecting, evaluating and correcting steps;

and further comprising the steps for said each half-segment in the given block and said corresponding collinear half-segment in one of said adjacent blocks separated therebetween by said one of said boundaries of;

(e) evaluating a dummy brightness level at said one boundary;

(1) equal to a sum of said corrected brightness level of said bordering pixel of said each half-segment and of a half difference of said corrected brightness level of said bordering pixel and a brightness level of a pixel of said corresponding half-segment which is adjacent to said bordering pixel when said corrected brightness level of said bordering pixel is more than said brightness level of said pixel adjacent to said bordering pixel, and

(2) equal to a difference of said corrected brightness level of said bordering pixel and said half-difference when said corrected brightness level of said bordering pixel is less than said brightness level of said pixel adjacent to said bordering pixel;

(f) evaluating a block divergence equal to a difference between two dummy brightness levels evaluated respectively for said each half-segment and said corresponding half-segment;

(g) determining a block correction factor as a function of said block divergence and of quantization pitches used in the coding and decoding by transformation of said given block and one of the adjacent blocks respectively.; and

(h) correcting brightness levels of pixel of said each half-segment and said corresponding half-segment, respectively;

(1) by subtracting from said pixel brightness levels said segment correction factor weighted by decreasing coefficients respectively in proportion to distances between said pixels and said boundary between said each half-segment and said corresponding half-segment, when said dummy brightness level relating to said each half-segment is more than said dummy brightness level relating to said corresponding half-segment, and

(2) by adding to said pixel brightness levels said block correction factor weighted by said coefficients when said dummy brightness level relating to said each half-segment is less than said dummy brightness level relating to said corresponding half-segment.

2. The method as claimed in claim 1 wherein said discontinuity correction factor is a substantially trapezoidal function of said discontinuity divergence for a predetermined quantization pitch.

3. The method as claimed in claim 1, wherein the integer N is equal to eight and said coefficients are 7/16, 5/16, 3/16 and 1/16.

4. The method as claimed in claim 1, wherein said segments are alternatively lines and columns of pixels in said blocks.

5. A digitalized image filtering device for filtering a digital decoded image obtained after encoding and decoding by transformation and quantization of blocks of pixels of an initial image of dimension N×

N, N being an integer, said digital decoded image being bound into a plurality of blocks each comprised of N parallel segments of pixels having respective brightness levels and being separated, in line and column, from adjacent decoded blocks by respective boundaries, comprising;

(a) input means for translating said decoded blocks according to a half-diagonal of said decoded blocks into translated blocks, each of said translated blocks having horizontal and vertical axes of symmetry corresponding to boundaries of said decoded blocks, thereby producing in series N pixel segments of each translated block, each of said pixel segments being formed with two collinear half-segments in two respective adjacent decoded blocks separated by a boundary;

(b) first evaluating means for successively evaluating, for each half-segment in said each of said pixel segments, a discontinuity divergence equal to an absolute value of a difference between a brightness level of a pixel of said each half-segment bordering on said boundary and an arithmetical average of brightness levels of two pixels adjacent to said bordering pixel and respectively included in said each half segment and a corresponding half-segment which borders on said boundary and is collinear with said each half-segment;

(c) first determining means for determining, for said each half segment, a discontinuity correction factor and a sign of said discontinuity divergence as a function of said discontinuity divergence and a quantization pitch used in said coding and decoding by transformation of the decoded block to which said half-segment belongs;

(d) first correcting means for correcting said brightness levels of said pixels in said each half-segment and corresponding collinear half-segment as a function of discontinuity correction factor and sign, thereby deriving first corrected brightness levels of said pixels in said each of said pixel segments and

(e) second evaluating means for successively evaluating, for said each of the pixel segments, a block divergence equal to a difference between two dummy brightness levels evaluated for said each half-segment and said corresponding collinear half segment, respectively, a dummy brightness level for each said half-segment being evaluated as a sum and a difference of the corrected brightness level of said bordering pixel and of a half-difference of said corrected brightness level and a brightness level of the pixel included in said corresponding collinear half-segment and bordering on said boundary when said corrected brightness level of said bordering pixel is less and more than said brightness level of said pixel included in said corresponding collinear half-segment, respectively;

(f) second determining means for determining a block correction factor and a sign of said block divergence as a function of said block divergence and said quantization pitch; and

(g) second correcting means for correcting brightness levels of pixels of said each half segment and said corresponding collinear half segment by adding and subtracting from said pixel brightness levels, respectively said block correction factor weighted by decreasing coefficients, in proportion to distances between said pixels and said boundary respectively when said dummy brightness level relating to said each half-segment is more and less than said dummy brightness level relating to said corresponding collinear half-segment.

6. The filtering device as claimed in claim 5, wherein said first and second determining means are merged, and further including:

(h) first multiplexing means for multiplexing said discontinuity divergences and block divergences into multiplexed divergences applied to said merged determining means which transmit multiplexed discontinuity and block correction factors to said first and second correcting means.

7. A filtering device as claimed in claim 5, wherein said first evaluating, determining and correcting means define a first means set in relation with first segments constituted by the lines and columns, respectively, in said translated blocks, and further wherein second evaluating, determining and correcting means define a second means set analogous to said first means set and relating to second segments constituted by columns in said translated blocks, and lines, respectively, said second means set operating subsequent to said first means set.

8. The device as claimed in claim 7, wherein said first and second divergence evaluating means are merged, and said first and second means for determining correction factors are merged, and further including:

(h) first transposing means for transposing each block of first corrected segments coming from said first means set into transposed blocks

(i) second multiplexing means for multiplexing said translated blocks and said transposed blocks into multiplexed blocks applied to said second means set; and

(j) second transposing means for transposing the corrected blocks produced by said second means set.