Statistically lossless compression system and method
AssigneeHarris Corporation
InventorsBruce Harold Pillman, Wayne Prentice, Michael E Napoli
Life of this patent
Estimating the term…
Abstract
A statistically lossless transform (SLT) can be used with data (e.g., image data) to be compressed. Prior to compression, an SLT stabilizes the variance for a read noise+Poisson process. The SLT redistributes (image) data that has been quantized with a linear uniform quantizer and essentially assures there are no gaps in the output values (i.e., all output digital values are used). The SLT re-quantizes the data with a quantization interval width that is proportional to the standard deviation of the process.
Claims20
In an encoding system including a data source, pre-processing circuitry, and a compressor, the improvement comprising:
- pre-processing circuitry configured to receive digital data from the data source, perform a statistically lossless transform on the received data, and output transformed data to the compressor, wherein the statistically lossless transform receives a series of m input code words and outputs a series of n output code words wherein m>n and wherein, for each of the input code words, a number of the adjacent input code words which are pre-processed to output a same one of the output code words is substantially proportional to a standard deviation of the adjacent input code words which are pre-processed to output the same one of the output code words, and wherein each input code word of the input code words is assigned at most one of the output code words.
A method of building lookup table for performing a statistically lossless transform, the method comprising the steps of:
- assigning output code words to which input code words of a sampled signal are to be transformed by calculating bin widths for the output code words, wherein the bin width of each of the output code words is substantially proportional to the standard deviation of a sampling of the input code words that are to be assigned the each of the output code words;
- constraining bin widths of the output code words to be at least a bin width of one;
- and
- storing in the lookup table the output code words indexed by the input code words.
Dependencies are inferred from the claim text.