Universal lossy compression methods
View Patent ↗The present invention provides methods for universal lossy compression that provide performance at or near the rate-distortion limit and that are based on universal, implementable lossy source coding algorithms.
1. A lossy data compression method, the method comprising:
receiving data from a source alphabet to be encoded, the source alphabet including one or more symbols; assigning coded bits to each symbol of the received data based on a probability distribution, wherein the probability distribution is obtained by sampling a sequence for reconstruction from a Boltzmann distribution associated with an energy function, the energy function depending on distortion between the source alphabet and the reconstruction; the energy function further depending on the reconstruction being compressible and on the desired distortion between the source alphabet and the reconstruction.
2. The method of claim 1 , further comprising: receiving compressed data, the compressed data including the coded bits representing the received data; and decoding the compressed data to recover and reconstruct each symbol of the received data based on the probability distribution.
3. The method of claim 1 , wherein assigning coded bits to each symbol of the received data based on a probability distribution comprises, at every iteration, the selection of an index I and replacement of the i-th sequence component.
4. The method of claim 1 , wherein the source alphabet is of small size.
5. The method of claim 1 , wherein the source alphabet is binary.
6. The method of claim 1 , wherein the lossy compression is sought for data for which no statistical characteristics are known.
7. The method of claim 1 , wherein the receiving/received data are still images.
8. The method of claim 1 , wherein the receiving/received data are images in motion including video and movies.
9. The method of claim 1 , wherein the receiving/received data are sounds including speech, voice and audio.