Transform-coding image compression for information efficiency and restoration

In this paper, we develop an image compression algorithm based on an information-theoretic analysis of transform coding in the end-to-end imaging process. Our analysis accounts for the radiance-field statistics, the image-gathering system design, and transform coding and compression in determining the information density of the transmitted image. This approach allows the specification of imaging systems that implement transform-coding image compression at arbitrary bit rates with maximum information density. Ultimately, the fidelity of an optimally restored image is limited by the information capacity of the imaging system. Our results indicate that with informationally optimized transform coding, even images compressed to low bit rates can be restored with relatively high fidelity.