EMD, which is short for Empirical Mode Decomposition, has been universally utilized in many fields of signal processing. A simplified algorithm of image compression with bi-dimensional EMD from open literature of EI and SPIE has been implemented and applied to compression of natural image, texture image and sonar image in this thesis.A non-uniform overlapping sampling using 7x7 block is applied in the thesis. The distribution of extremes steers the sample rate of each block, with the sampling-points block being the result. The sampling rates of rows and columns are identical with each other, it being determined by the higher one of local frequencies from two directions. It can be seen that the quality of reconstructed image weakens with the increasing of threshold, yet this conclusion from natural and texture image is not the same with the one from sonar image. This phenomenon is analyzed briefly.Also, coding of the EMD with DCT of the non-uniform sampled blocks (NSEMDCT) and mode-image coding (MI) have been implemented and been applied to image compression. |