Pyramid data structures for an on-line image progressive transmission

Pyramid data structures are considered to be an important class of such image representations. A pyramid consists of several levels of reduced-sized image each correspond to successive approximation of the original image. This structure is very suitable for progressive transmission and requires relatively low computations. The computation is local and the same procedure is iterated to build each pyramidal level from its predecessors. Progressive transmission will be achieved by sending the pyramid starting from top level (low resolution) to the bottom level (high resolution). In this paper, we introduce a new algorithm for pyramid data structure generation. This new algorithm is based on the subsampling approach rather the averaging approach. The resulting structure requires minimal time to be generated. The subsampling approach may be followed as many images are still recognized even with a subsampling rate in both directions. We follow this direction and propose the selection rule pyramid. However, this pyramid structure will be very suitable with images which contain dominant low-frequency components. But, in texture images, it results in a bad approximations from the subjective point of view. To obtain efficiency, we generate the selection-difference pyramid from the selection rule pyramid, each node in a certain level will be subtracted from its sibling in the predecessor level. Wang & Goldberg recently reviewed various pyramid data structure. They demonstrate that the reduced-difference pyramid achieves the best performance in terms of three criteria: the equivalent entropy, rate distortion performance and total lossless transmission bit rate. We compare the proposed structure to other structures, according to Wang and Goldberg criteria, we arrived to better performance. Simulation are carried out over group of images. Results and discussion are listed.