Method for rectifying channel errors in a transmitted image signal encoded by classified vector quantization
First Claim
1. A method, for use in an image signal decoding system, for concealing class errors in an image signal transmitted in a compressed form by using a classified vector quantization method, wherein the image signal is divided into a multiplicity of blocks, each of said blocks being classified into one of a plurality of classes based on an edge pattern of said each block and said classes including a non-edge class which corresponds to a block having no edge pattern therein and a texture class which corresponds to a block having more than one edge patterns therein, said method comprising the steps of:
- detecting a block containing a class error;
detecting boundary edges at each of four groups of neighboring pixels to the class error containing block, wherein each group of the neighboring pixels is located at a boundary of the class error containing block and included in one of four contiguous blocks to the class error containing block, pixel values of the neighboring pixels being considered to be correct;
determining the non-edge class as a class of the class error containing block is less than two boundary edges are detected and the texture class, if more than two boundary edges are detected;
obtaining an edge pattern in the class error containing block by connecting the detected boundary edges if two boundary edges are detected;
determining the class for the class error containing block by selecting the most similar edge pattern to the obtained edge one, wherein said class is one of said plurality of classes; and
repeating the above steps until classes all the blocks containing class errors are determined.
1 Assignment
0 Petitions
Accused Products
Abstract
A novel method for concealing or rectifying channel errors present in a decoded image signal which has been transmitted in a compressed form by using a classified vector quantization technique comprises a step for estimating the class for a current block by detecting boundary edges at each of four groups of neighboring pixels; and a step for estimating the representative vector for the current block by calculating a multiple number of side-matching functions and choosing the candidate vector producing the smallest side-matching function as the representative vector for the current block.
150 Citations
2 Claims
-
1. A method, for use in an image signal decoding system, for concealing class errors in an image signal transmitted in a compressed form by using a classified vector quantization method, wherein the image signal is divided into a multiplicity of blocks, each of said blocks being classified into one of a plurality of classes based on an edge pattern of said each block and said classes including a non-edge class which corresponds to a block having no edge pattern therein and a texture class which corresponds to a block having more than one edge patterns therein, said method comprising the steps of:
-
detecting a block containing a class error; detecting boundary edges at each of four groups of neighboring pixels to the class error containing block, wherein each group of the neighboring pixels is located at a boundary of the class error containing block and included in one of four contiguous blocks to the class error containing block, pixel values of the neighboring pixels being considered to be correct; determining the non-edge class as a class of the class error containing block is less than two boundary edges are detected and the texture class, if more than two boundary edges are detected; obtaining an edge pattern in the class error containing block by connecting the detected boundary edges if two boundary edges are detected; determining the class for the class error containing block by selecting the most similar edge pattern to the obtained edge one, wherein said class is one of said plurality of classes; and repeating the above steps until classes all the blocks containing class errors are determined.
-
-
2. A method, for use in an image signal decoding system, for concealing channel errors in an image signal transmitted in a compressed form by using a classified vector quantization method, wherein the image signal is divided into a multiplicity of blocks, each of said blocks being classified into one of a plurality of classes based on an edge pattern of said each block, said classes including a non-edge class which corresponds to a block having no edge pattern therein and a texture class which corresponds to a block having more than one edge patterns therein, each of said classes corresponding to a codebook including a multiple number of candidate vectors, and each of the blocks included in the same class being represented by a representative vector selected from the candidate vectors included in the codebook corresponding to the class, said method comprising the steps of:
-
detecting a block containing a channel error; detecting boundary edges at each of four groups of neighboring pixels to the channel error containing block, wherein each group of the neighboring pixels is located an a boundary of the channel error containing block and included in one of four contiguous blocks to the channel error containing block, pixel values of the neighboring pixels being considering to be correct; determining the non-edge class as a class of the channel error containing block if less than two boundary edges are detected and the texture class, if more than two boundary edges are detected; obtaining an edge pattern in the channel error containing block by connecting the detected boundary edges if two boundary edges are detected; determining the class for the channel error containing block by selecting the most similar edge pattern to the obtained edge one, wherein said class is one of said plurality of classes; calculating a multiple number of side-matching functions, each of said functions representing the difference between the boundary pixels of the channel error containing block and the neighboring pixels to the channel error containing block, wherein each of said side-matching functions is calculated for each of the multiple number of candidate vectors include in the codebook corresponding to the class where the channel error containing block belongs; choosing the candidate vector producing the smallest side-matching function as the representative vector for the channel error containing block; and repeating the above steps until classes and representative vectors for all the blocks containing channel errors are determined.
-
Specification