Lossy plus lossless techniques for image compression split an image into a low-bit-rate lossy representation and a residual that represents the difference between this low-rate lossy image and the original. Conventional schemes encode the lossy image and its lossless residual in an independent manner. We show that making use of the lossy image to encode the residual can lead to significant savings in bit rate. Further, the complexity increase to attain these savings is minimal. The savings are achieved by capturing the inherent structure of the image in the form of a noncausal prediction model that we call a prediction tree. This prediction model is then used to transmit the lossless residual. Simulation results show that a reduction of 0.5 to 1.0 bit/pixel can be achieved in bit rates compared to the conventional approach of independently encoding the residual.
ASJC Scopus subject areas
- Atomic and Molecular Physics, and Optics
- Computer Science Applications
- Electrical and Electronic Engineering