Method for detecting class errors in a transmitted image signal encoded by classified vector quantization

Claim 1

A method for detecting an error in a class name of a current block for use in an image decoding system for decoding an image signal encoded by using a classified vector quantization method wherein an image frame of the image signal is divided into a multiplicity of blocks, each of said blocks being classified into one of (M+1) classes based on an edge pattern of said each of the blocks, said (M+1) classes including a non-edged class having no edge therein and M edged classes, each of the M edged classes having an edge therein with M being an integer greater than 1, and each of said (M+1) classes being given a class name, said method comprising the steps of: decoding the image signal to provide a class name and a cordword index for each of the blocks; generating a reconstructed block of pixel values for each of the blocks by dequantizing the cordword index through the use of a cordbook selected in response to the class name of said each of the blocks; storing reconstructed blocks of pixel values to form a reconstructed image frame signal; determining, based on a decoded class name of the current block, whether the current block belongs to one of the M edged classes or corresponds to the non-edged class and selecting reference blocks out of neighboring blocks of the current block; and if the current block belongs to one of the edged classes, determining the class error of the current block based on a degree of similarity between an edge pattern represented by the decoded class name of the current block and an edge pattern represented by a decoded class name of each of the selected reference blocks and if the current block corresponds to the non-edged class, obtaining, from the reconstructed image frame signal, boundary pixel values of the current block and their neighboring pixel values within the selected reference blocks to thereby determine the class error of the current block based on the obtained pixel values.

Claim 2

A method for detecting an error in a class name of a current block for use in an image decoding system for decoding an image signal encoded by using a classified vector quantization method wherein an image frame of the image signal is divided into a multiplicity of blocks, each of said blocks being classified into one of (M+1) classes based on an edge pattern of said each of the blocks, said (M+1) classes including a non-edged class having no edge therein and M edged classes, each of the M edged classes having an edge therein with M being an integer greater than 1, and each of said (M+1) classes being given a class name, said method comprising the steps of: decoding the image signal to provide a class name and a cordword index for each of the blocks; generating a reconstructed block of pixel values for each of the blocks by dequantizing the cordword index through the use of a cordbook selected in response to the class name of said each of the blocks; storing reconstructed blocks of pixel values to form a reconstructed image frame signal; determining, based on a decoded class name of the current block, whether the current block belongs to one of the M edged classes or corresponds to the non-edged class and selecting reference blocks out of neighboring blocks of the current block; and if the current block belongs to one of the edged classes, determining the class error of the current block based on a degree of similarity between an edge pattern represented by the decoded class name of the current block and an edge pattern represented by a decoded class name of each of the selected reference blocks and if the current block corresponds to the non-edged class, obtaining, from the reconstructed image frame signal, boundary pixel values of the current block and their neighboring pixel values within the selected reference blocks to thereby determine the class error of the current block based on the obtained pixel values.