Process for changing the resolution of MPEG bitstreams, and a system and a computer program product therefor

A system for generating an MPEG output bitstream starting from an MPEG input bitstream having a first resolution, said MPEG output bitstream having a second resolution modified with respect to the first resolution of said MPEG input bitstream, the system comprising:

• a sorting module configured to distinguish, in said input bitstream, first portions that substantially do not affect and second portions that substantially do affect variation of resolution of the MPEG output bitstream;
• and
• an inverse quantizer module structured to subject said second portions of the input bitstream to an inverse-quantization operation to obtain inverse-quantized second portions of the first resolution;
• a motion compensation module structured to subject said inverse-quantized second portions of the first resolution to motion compensation in a DCT domain to obtain motion-compensated second portions of the first resolution;
• at least one processing module configured to subject said motion-compensated second portions to a function of modification of the resolution obtained by filtering said motion-compensated second portions in the discrete cosine transform domain to obtain resolution-modified second portions in the second resolution, wherein said at least one processing module is further configured to:
• store a given number of macroblocks aligned on one and the same line;
• and
• multiply said macroblocks by at least one matrix with a scaled factor of definition, wherein the multiplying operation comprises:
• multiplying a first subset of macroblocks of said given number of macroblocks by a first matrix with a factor of definition that is reduced in a first direction to obtain a first macroblock having the factor of definition that is reduced in the first direction;
• multiplying a second subset of macroblocks of said given number of macroblocks by a second matrix with the factor of definition that is reduced in the first direction to obtain a second macroblock the factor of definition that is reduced in the first direction;
• and
• multiplying the first and second macroblocks by a third matrix with a factor of definition that is reduced in a second direction to obtain a third macroblock;
• at least one storage module configured to store said resolution-modified second portions;
• and
• an upsample filter configured to obtain resolution-remodified second portions by subjecting the stored resolution-modified second portions to a second operation of modification of the resolution that brings back the resolution of the stored data from said second resolution to said first resolution, wherein the motion compensation module is structured to obtain the motion-compensated second portions of the first resolution using the resolution-remodified second portions as a prediction signal.

A computer program product directly loadable in the memory of a digital computer and comprising software code portions for causing a computer to generate an output bitstream from an input bitstream having a first resolution, the output bitstream having a second resolution modified with respect to the first resolution of the input bitstream, by:

• distinguishing, in the input bitstream, first portions that substantially do not affect and second portions that substantially do affect variation of the resolution of the output bitstream;
• subjecting said second portions of the input bitstream to an inverse-quantization operation to obtain inverse-quantized second portions of the first resolution;
• subjecting said inverse-quantized second portions of the first resolution to motion compensation in a DCT domain to obtain motion-compensated second portions of the first resolution;
• subjecting the motion-compensated second portions to a function of modification of the resolution obtained by filtering the motion-compensated second portions in a discrete cosine transform domain to obtain resolution-modified second portions in the second resolution, wherein the filtering in the domain of the discrete cosine transform includes:
• store a given number of macroblocks aligned on one and the same line;
• and
• multiply said macroblocks by at least one matrix with a scaled factor of definition, wherein the multiplying operation comprises:
• multiplying a first subset of macroblocks of said given number of macroblocks by a first matrix with a factor of definition that is reduced in a first direction to obtain a first macroblock having the factor of definition that is reduced in the first direction;
• multiplying a second subset of macroblocks of said given number of macroblocks by a second matrix with the factor of definition that is reduced in the first direction to obtain a second macroblock the factor of definition that is reduced in the first direction;
• and
• multiplying the first and second macroblocks by a third matrix with a factor of definition that is reduced in a second direction to obtain a third macroblock;
• storing said resolution-modified second portions;
• and
• obtaining resolution-remodified second portions by subjecting the stored resolution-modified second portions to a second operation of modification of the resolution that brings back the resolution of the stored data from said second resolution to said first resolution, wherein subjecting said inverse-quantized second portions of the first resolution to motion compensation in the DCT domain includes using the resolution-remodified second portions as a prediction signal when motion compensating said inverse-quantized second portions of the first resolution.