Constrained discrete-cosine-transform coefficients for better error detection in a corrupted MPEG-4 bitstreams

An error-detecting decoder comprising:

• a bitstream parser coupled to receive compressed video data and locate a block of pixels within the compressed video data and output compressed block data;
• a variable-length decoder, receiving the compressed block data and converting variable-length codewords in the compressed block data to block coefficients;
• a constraint checker, receiving the block coefficients, for selecting at least two coefficients in the block coefficients and signaling an error when the at least two coefficients fail to meet a constraint;
• an error concealer, activated by an error detected by the constraint checker, for preventing the block coefficients from being used to generate pixels for display when the error is detected;
• and
• a coefficient transformer, receiving the block coefficients, for generating block data by inverse transforming the block coefficients, the block data being either pixels for display in the block or error terms for generating pixels for display for the block from stored pixels,
• whereby errors in the block are detected and concealed when coefficients fail to meet the constraint.

A compressor for adjusting a video stream for error-detection comprising:

• discrete cosine transform (DCT) means, receiving blocks of pixels or error terms for blocks of pixels, for discrete-cosine transforming a block of pixels or error terms to generate DCT coefficients;
• quantize means, receiving the DCT coefficients, for discarding least-significant bits of low-valued DCT coefficients to generate zero-valued DCT coefficients;
• constraint-adjust means, receiving DCT coefficients, for adjusting values of DCT coefficients to embed error-detection information into adjusted DCT coefficients;
• and
• variable-length encode means, receiving the adjusted DCT coefficients, for coding the adjusted DCT coefficients as codewords of varying lengths, the codewords requiring less storage capacity than the DCT coefficients for an average block in the video stream;
• wherein the constraint-adjust means comprises:
• select means for selecting at least two coefficients from the DCT coefficients;
• average means, receiving the at least two coefficients from the select means, for generating an average magnitude from magnitudes of the at least two coefficients;
• and
• substitute means for replacing magnitudes of the at least two coefficients with the average magnitude without altering signs of the at least two coefficients, and for not altering other DCT coefficients to generate the adjusted DCT coefficients,
• whereby magnitudes are averaged and substituted and whereby the video stream is embedded with error-detection information and compressed.