Detecting block artifacts in coded images and video
View Patent ↗A method is provided for block artifact detection and block artifact determination using multiple features in coded images and video. A method includes identifying a plurality of overlapping blocks in a picture with respect to a current block. Each of the plurality of overlapping blocks covers a portion of the current block and a portion of an adjacent block. The method further includes transforming at least one of the plurality of overlapping blocks into respective coefficients corresponding to at least one of a luminance component and a chrominance component. The method also includes identifying a block artifact for the at least one of the plurality of overlapping blocks when coefficients corresponding to the at least one of the plurality of overlapping blocks are larger than a predefined value.
1. A method for detecting block artifacts, comprising:
identifying a plurality of overlapping blocks in a picture from a sequence of moving video pictures with respect to a current block, where each one of said plurality of overlapping blocks covers a portion of the current block and a portion of an adjacent block;
transforming at least one of the plurality of overlapping blocks into respective transform coefficients corresponding to at least one of a luminance component and a chrominance component;
identifying a block artifact for the at least one of the plurality of overlapping blocks when a coefficient of said respective transform coefficients corresponding to the at least one of the plurality of overlapping blocks is larger than a value;
determining a blockiness value for a block boundary between a first block and a second block in the picture, including:
accessing lines of pixels, each of the lines of pixels having pixels from the first block and the second block,
determining a respective filtered value for each of the lines of pixels, and
determining the blockiness value for the block boundary in response to the filtered values, the determining the blockiness value excluding a minimum or a maximum of the filtered values; and
determining an artifact strength of the block artifact responsive to at least one of the identifying a block artifact and the determined blockiness value.
2. The method of claim 1 , wherein said step of identifying the block artifact uses at least one threshold as said value to determine whether the block artifact exists within said at least one of the overlapping blocks and a direction of the block artifact with respect to the current block.
3. The method of claim 1 , further comprising generating an overall blockiness metric for the picture based on a number of overlapping blocks in the picture that have the block artifact.
4. The method of claim 3 , wherein said number of overlapping blocks is determined from a total number of overlapping blocks in said picture that have the block artifact.
5. The method of claim 1 , further comprising generating a blockiness artifact map that specifies a location of the block artifact in the picture.
6. The method of claim 5 , further comprising:
performing feature detection on a reconstructed version of the picture; and
performing feature based artifact correction on the blockiness artifact map to generate a feature-enhanced blockiness artifact map.
7. The method of claim 1 , wherein said step of identifying the block artifact further comprises the step of eliminating a false identification of the block artifact.
8. The method of claim 7 , wherein said eliminating step is performed when an adjusted and filtered pixel difference at a block boundary corresponding to the identified block artifact in a reconstructed version of the picture is less than, by an amount, an adjusted filtered pixel difference in the block boundary corresponding to the identified block artifact in an original version of the picture.
9. The method of claim 7 , wherein said eliminating step is performed when the transform coefficients of a particular one of the plurality of overlapping blocks in a reconstructed version of the picture is less than, by an amount, the transform coefficients of a co-located block for the particular one of the plurality of overlapping blocks in an original version of the picture.
10. The method of claim 1 , further comprising calculating an overall picture level block artifact detection value for the picture based on a weighted result of said blockiness value for the block boundary, wherein a weighting factor is determined based on at least one detected feature in the picture.
11. The method of claim 10 , wherein the feature detection involves at least one of the following features comprising: brightness, chromaticity, skin color, texture, location, and neighborhood information.
12. An apparatus for detecting block artifacts, comprising:
a transform domain artifact detector configured to
identify a plurality of overlapping blocks in a picture from a sequence of moving video pictures with respect to a current block, where each one of said plurality of overlapping blocks covers a portion of the current block and a portion of an adjacent block,
transform at least one of the plurality of overlapping blocks into respective transform coefficients corresponding to at least one of a luminance component and a chrominance component, and
identify a block artifact for the at least one of the plurality of overlapping blocks when a coefficient of said respective transform coefficients corresponding to the at least one of the plurality of overlapping blocks is larger than a value;
a pixel domain artifact detector for determining a blockiness value for a block boundary between a first block and a second block in the picture configured to
access lines of pixels, each of the lines of pixels having pixels from the first block and the second block,
determine a respective filtered value for each of the lines of pixels, and
determine a blockiness value for the block boundary in response to the filtered values, excluding a minimum or a maximum of the filtered values; and
an artifact strength determiner and metric assigner configured to calculate an artifact strength of the block artifact responsive to at least one of the identification of the block artifact and the determined blockiness value.
13. The apparatus of claim 12 , wherein the transform domain artifact detector uses at least one threshold as said value to determine whether the block artifact exists within said at least one of the overlapping blocks and a direction of the block artifact with respect to the current block.
14. The apparatus of claim 12 , wherein the artifact strength determiner and metric assigner generates an overall blockiness metric for the picture based on a number of overlapping blocks in the picture that have the block artifact.
15. The apparatus of claim 14 , wherein said number of overlapping blocks is determined from a total number of overlapping blocks in said picture that have the block artifact.
16. The apparatus of claim 12 , further comprising a merger module configured to generate a blockiness artifact map that specifies a location of the block artifact in the picture.
17. The apparatus of claim 16 , further comprising:
a feature detector configured to perform feature detection on a reconstructed version of the picture; and
a feature based artifact classifier configured to perform feature based artifact correction on the blockiness artifact map to generate a feature-enhanced blockiness artifact map.
18. The apparatus of claim 12 , further comprising:
an artifact map refiner configured to eliminate a false identification of the block artifact.
19. The apparatus of claim 18 , wherein said artifact map refiner is configured to eliminate the false identification of the block artifact when an adjusted and filtered pixel difference at a block boundary corresponding to the identified block artifact in a reconstructed version of the picture is less than, by an amount, an adjusted filtered pixel difference in the block boundary corresponding to the identified block artifact in an original version of the picture.
20. The apparatus of claim 18 , wherein said artifact map refiner is configured to eliminate the false identification of the block artifact when the transform coefficients of a particular one of the plurality of overlapping blocks in a reconstructed version of the picture is less than, by an amount, the transform coefficients of a co-located block for the particular one of the plurality of overlapping blocks in an original version of the picture.
21. The apparatus of claim 12 , wherein said artifact strength determiner and metric assigner are configured to calculate an overall picture level block artifact detection value for the picture based on a weighted result of said blockiness value for the block boundary, wherein a weighting factor is determined based on at least one detected feature in the picture.
22. The apparatus of claim 21 , wherein the feature detection involves at least one of the following features comprising: brightness, chromaticity, skin color, texture, location, and neighborhood information.