Methods for compression using a denoiser
View Patent ↗Various embodiments of the present invention provide a compression method and system that compresses received data by first denoising the data and then losslessly compressing the denoised data. Denoising removes high entropy features of the data to produce lower entropy, denoised data that can be efficiently compressed by a lossless compression technique. One embodiment of the invention is a universal lossy compression method obtained by cascading a denoising technique with a universal lossless compression method. Alternative embodiments include methods obtained by cascading a denoising technique with one or more lossy or lossless compression methods.
1. A method for compressing data comprising:
receiving the data;
repeatedly denoising the data using a denoiser to produce denoised data, each time selecting different denoising parameters wherein denoising parameters comprise
symbol compositions of contexts within which symbols are evaluated by the denoiser;
lengths of the contexts within which symbols are evaluated by the denoiser;
values stored in a channel-transition matrix Π; and
values stored in a distortion matrix Λ, until a distortion computed for the denoised data is lower than a threshold distortion; and
compressing the denoised data to produce compressed data.
2. The method of claim 1 wherein denoising and compressing are repeated until compressed data with a size lower than a threshold size is produced.
3. The method of claim 1 wherein repeatedly denoising the data using a denoiser to produce denoised data, each time selecting different denoising parameters, until a distortion computed for the denoised data is lower than a threshold distortion is carried out by one of:
an automated system;
a semi-automated system that includes user interaction; and
a manually controlled system in which a user interactively selects and adjusts denoising parameters.
4. The method of claim 1 wherein the denoiser is tuned by modifying the channel-transition matrix Π to find an optimal channel transition matrix Π that produces an optimized denoised signal leading to greatest compressibility and acceptable distortion.
5. The method of claim 4 further including:
setting entries of each row i of the channel-transition matrix Π, denoted π i,1 , . . . , π i,n to be a probability distribution having a largest entropy subject to a constraint that
∑
j
π
i
,
j
λ
i
,
j
<
Δ
where λ i,j are entries of the ith row of the distortion matrix Λ and Δ is a threshold distortion level.
6. The method of claim 5 wherein an initial value for Δ is selected as a desired distortion level between lossy representation of the data and the uncompressed data.
7. The method of claim 5 wherein Δ is iteratively adjusted as a denoising parameter in order to converge on a desired ratio between compression efficiency and signal distortion.
8. The method of claim 1 wherein compressing the data further includes compressing the data using one of:
a lossy decompression method; and
a lossless decompression method.
9. The method of claim 1 further including subsequently decompressing the compressed data by reversing the compression using one of:
a lossy decompression method; and
a lossless decompression method.
10. A computer readable storage medium containing executable instructions which, when executed in a processing system, causes the system to perform a method for compressing data comprising:
receiving the data;
repeatedly denoising the data using a denoiser to produce denoised data, each time selecting different denoising parameters wherein denoising parameters comprise
symbol compositions of contexts within which symbols are evaluated by the denoiser;
lengths of the contexts within which symbols arc evaluated by the denoiser;
values stored in a channel-transition matrix Π; and
values stored in a distortion matrix Λ, until a distortion computed for the denoised data is lower than a threshold distortion; and
compressing the denoised data to produce compressed data.
11. The computer readable storage medium of claim 10 wherein denoising and compressing are repeated until compressed data with a size lower than a threshold size is produced.
12. The computer readable storage medium of claim 10 wherein repeatedly denoising the data using a denoiser to produce denoised data, each time selecting different denoising parameters, until a distortion computed for the denoised data is lower than a threshold distortion is carried out by one of:
an automated system;
a semi-automated system that includes user interaction; and
a manually controlled system in which a user interactively selects and adjusts denoising parameters.
13. The computer readable storage medium of claim 10 wherein the denoiser is tuned by modifying the channel-transition matrix Π to find an optimal channel transition matrix Π that produces an optimized denoised signal leading to greatest compressibility and acceptable distortion.
14. The computer readable storage medium of claim 13 wherein the method further includes:
setting entries of each row i of the channel-transition matrix Π, denoted π i,1 , . . . , π i,n to be a probability distribution having a largest entropy subject to a constraint that
∑
j
π
i
,
j
λ
i
,
j
<
Δ
where λ i,j are entries of the ith row of the distortion matrix Λ and Δ is a threshold distortion level.
15. The computer readable storage medium of claim 14 wherein an initial value for Δ is selected as a desired distortion level between lossy representation of the data and the uncompressed data.
16. The computer readable storage medium of claim 14 wherein Δ is iteratively adjusted as a denoising parameter in order to converge on a desired ratio between compression efficiency and signal distortion.
17. The computer readable storage medium of claim 10 wherein compressing the data further includes compressing the data using one of:
a lossy decompression method; and
a lossless decompression method.
18. The computer readable storage medium of claim 10 wherein the method further includes subsequently decompressing the compressed data by reversing the compression using one of:
a lossy decompression method; and
a lossless decompression method.