Novel contrast pyramid coding of images

Abstract

Different pyramid coding methods have been proposed, but they a common property in that a difference image is taken as the information difference image, such as an image in the Laplacian pyramid. We propose a new pyramid structure instead of the difference pyramid structure to represent the information difference between two adjacent level images. The proposed pyramid is obtained from the expanded higher level image and the current level image with a simple nonlinear algorithm. The value of the image in the new pyramid is close related with a local contrast measure, and it is thus called a contrast pyramid. Both theoretical analysis and simulation results have shown that with the proposed contrast pyramid structure, amplitude redundancy can be easily exploited based on the human visual phenomena, such as Weber's law and the spatial masking effect.