Image Compression Algorithm Based On Empirical Mode Decomposition

Empirical Mode Decomposition method proposed by Norden E. Huang of NASA is a new processing method for non-stationary signal. As a new signal analytical approach, EMD method decomposes a complex signal into a series of Intrinsic Mode Functions and remaining final trend by the signal itself.Bi-dimensional Empirical Mode Decomposition method which is the expansion of the EMD method in the two-dimensional field is a new analytical method for image multi-scale analysis and only has a short-time development. In recent years, BEMD method has been used in the field of image texture analysis, image denoising, image fusion and so on,. But it has little research on the image coding area. The characteristic of BEMD in image coding and compression is researched in the paper. The combination of these two fields not only pours into new vigor for the mature technology of image coding and compression, but also expands the development of the BEMD method in the application field. So it has some research value and innovation.On the basis of introducing the three kinds of analytical method of non-stationary signal briefly which are Short-time Fourier Transform, the Wigner-Ville Distribution and Wavelet analysis, the EMD method, especially its expansion in two-dimensional field, is discussed in detail in this paper. Against the shortcomings of existing BEMD algorithm such as slow decomposition and high time complexity, extrema point locating, surface interpolation and termination criteria are improved in this paper. Under the premise of protecting decomposition effect, the decomposition rate of the modified BEMD algorithm is accelerated and the time is saved. The existing image compression algorithms based on BEMD took a unified coding strategy to process IMF and surplus image, which didn't take full use of the different characteristics of the two. According to the different characteristic of IMF and surplus image, a kind of classified compression algorithm based on BEMD which adopts different strategy to IMF and surplus image has been designed. A multi-resolution analytical method for image IMF is proposed. And then, an image compression algorithm based on this method is designed. Finally, experimental results show that whether the compression ratio or the PSNR, the algorithm proposed by this paper is superior to the traditional JPEG method.