Wavelet Transform In Image Processing Applications

Wavelet analysis is a tool of time-frequency analysis after Fourier analysis. In the field of image processing, its application covered imaging technique, image pre-processing, image compression and transferring, image registration, image analysis, feature extraction and pattern classification, etc. In this paper, it's researched on wavelets application in the fields of multi-scale edge detection, remote sensing image processing and medical imaging.The traditional methods of edge detection are based on one-order derivative's maximum, or two-order derivative's zero-crossing. This kind of edge definition is very sensitive to noises. And thus, edge detection should be carried out in large scale, by which the image was smoothed. One of the shortcomings of edge detection in large scale is that it's difficult to locate edge precisely, which will make mistakes in pattern recognition based on edge features. With multi-scale characterization, wavelet analysis was widely used to multi-scale edge detection. In this paper, it was proved that, wavelet-based multi-scale edge detection would keep edge positions very well, if symmetric bases were used in wavelet transform. Furthermore, an algorithm of multi-scale edge detection based on bi-orthogonal symmetric wavelet was put forward, with which, "good edges" will be obtained while the edge positions will be kept well.As a new technique applying to protecting the copyright of digital productions, the digital watermark technique has drawn extensive attention. A method of embedding the watermark in digital images based on the discrete wavelet transform is proposed. The watermark used here is not the conventional patterns such as a pseudo-random sequence or a bit stream but a text watermark. The information which the text watermark contains is abundant and intuitionistic, also the watermark is robust. To ensure the security of the watermark and make the watermark be hard to be extracted, the watermark is scrambled with Arnold scrambling transformation before embedded into the original image. According to the different characteristics of the high and the low frequency components of the wavelet coefficients of the original image, more watermark information is embedded in the high frequency components while less information in the low ones. That is to say, by using the hierarchical structure of the wavelet, the watermark is repeatedly embedded in various places. Moreover, experimental results have proved that the method is robustenough to some image degradation process such as cropping, JPEG compression and sharpening etc.Embedded ZeroTree Wavelet Coding(EZW) algorithm, it is found that this algorithm does not make full use of the properties of Human Visual System(HVS) to eliminate visual redundancy. An improved image compression algorithm based on bi-orthogonal wavelet coefficients is thus presented, maily including: 1. The subimage of the lowest frequency(LL4) is carried out lossless compressed coding;2 The subimages on diagonal direction of the highest frequency(HHl) is abandoned and is not carried out coding, because it is of great probability for zero, and it little affects visual. 3 The rest of the subimages are given different bit numbers according to their different characteristics. Then they are carried out quantization of zerotree and run-length coding. Experimental results show that this algorithm is more effective.