Method of variable-length encoding of images and device for performing said
AssigneeU.S. Philips Corporation
InventorEtienne Fert
Life of this patent
Estimating the term…
Abstract
A method for variable-length encoding of images includes a quantization step followed by a variable-length encoding step, whereafter the signals obtained are stored in a buffer memory. The quantization step used is computed on the basis of a feedback parameter, which is a decreasing function of the filling level of the buffer memory, and a feedforward parameter, which is a function of the quantity of information in the image. Advantageously, the feedforward parameter includes at least a multiplicative factor which is a decreasing function of the saturation of one of the chrominance components U or V in the current block or macroblock with respect to the rest of the image. An encoding device is also described.
Claims21
A method of encoding images represented by digital signals organized in luminance and chrominance blocks grouped into macroblocks, comprising the steps of:
- quantizing the digital signals to produce quantized signals;
- variable-length encoding said quantized signals to produce encoded signals;
- and
- storing said encoded signals in a buffer memory,
- wherein a quantization step used in performing the quantizing step is computed on the basis of a feedback parameter which is a decreasing function of the filling level of the buffer memory, and a multiplicative correction factor in the form of a feedforward parameter which is a function of the quantity of information in the image currently being encoded, and
- wherein said feedforward parameter comprises at least one chrominance factor for a current block or macroblock, which is a decreasing function of the saturation of one of chrominance components U and V in the current block or macroblock, with respect to the rest of the image.
A device for encoding images represented by digital signals organized in luminance and chrominance blocks grouped into macroblocks, comprising:
- a first module for quantizing the digital signals to produced quantized signals;
- a second module for variable-length encoding said quantized signal to produce encoded signals;
- a buffer memory receiving and storing said encoded signals;
- and
- a third module disposed between said buffer memory and said first module for bitrate control of an output of said buffer memory, said third module including means for modifying a quantization step utilized by said first module responsive to both a feedback parameter, which is a decreasing function of a filling level of said buffer memory, and a multiplicative correction factor in the form of a feedforward parameter, which is a function of the quantity of information in the image currently being encoded and which is provided by a weighting module;
- wherein said weighting module comprises at least one module for estimating at least one chrominance factor for the current block or macroblock, which is a decreasing function of the saturation of one of the chrominance components U and V in the current block or macroblock with respect to the rest of the image currently being encoded.
A device for modifying a quantization step used in system for encoding images represented by digital signals organized in luminance and chrominance blocks grouped into macroblocks, the system having a first module for quantizing the signals, a second module for variable-length encoding the quantized signals, and a buffer memory receiving the output of said second module, the device comprising:
- a weighting module including at least one module for estimating a chrominance factor for the current block or macroblock, which is a decreasing function of the saturation of one of the chrominance components U and V in the current block or macroblock with respect to the rest of the image currently being encoded;
- and
- a control module disposed between the buffer memory and the first module for bitrate control of the output of the buffer memory, said control module including means for modifying a quantization step utilized by the first module responsive to both a feedback parameter, which is a decreasing function of a filling level of the buffer memory, and a feedforward parameter, which is a function of the quantity of information in the image currently being encoded.
Dependencies are inferred from the claim text.