Texture is an important characteristic for the analysis of many types of images. It can be seen in all images from multispectral scanner images obtained from aircraft or satellite platforms (which the remote sensing community analyzes) to microscopic images of cell cultures or tissue samples (which the biomedical community analyzes). In this dissertation, we investigate the method of bivariate texture-matched nonseparable wavelet filterbanks for hierarchical texture analysis. Adaptive signal modeling method is the typical representation of the new generation object-orient signal processing. It is the focus of today's signal modeling method. Texture-matched modeling method is an important aspect of Computer Vision, which has been received increased attention in the literature. The first part of this thesis provides a review of background materials related to texture modeling as well as a summary of traditional methods of texture analysis. The difficulties and the newly developments of texture analysis is discussed, which focus on the adaptive modeling method. The main goal of this thesis is the design of bivariate texture-matched nonseparable wavelet filterbanks for texture compression. Connecting with the methods of F.Ade and T.Greiner, we have developed a new method to design the 4×4 and 6×6 texture-matched nonseparable wavelet filterbanks for texture compression. A comparison of the results with the 4×4 Daubechies wavelet, Haar wavelet and Peng wavelet is described in Chapter 3. In addition, we generalize our method to n-dimension space for arbitrary filter-length. In the closing chapter, the key points of the thesis are summarized. The conclusion also discusses extensions to Computer Vision and provides suggestions for further work related to texture compression.
修改评论