Study Of Image/Video Compression And Watermarking Techniques Based On Wavelet Transform

In recent years, the study of wavelet transform urged its usage in image coding. The multiresolution analysis of wavelet transform matches with the character of human vision. It can give sub-bands for different levels and different directions of an image. Studying the application of wavelet transform in arbitrarily shaped image coding, 3-D transform coding of video, post-processing of decoded image, can promote the consummation of new image/video coding standard, etc., MPEG-4, JPEG-2000.The protection and enforcement of intellectual property rights has become an important issue in image/video coding system. Watermarking is the process of encoding hidden copyright information in an image by making small modifications to its pixels. It can be used for image authentication and forgery prevention. Studying watermark in wavelet domain of image/video, is helpful to include watermarking system into image/video coding standard based on wavelet transform.In this dissertation, following the development of wavelet theory, especially the second generation wavelet transform (the lifting algorithm), we study the usage of wavelet transform in arbitrarily shaped image coding, 3-D transform coding of video, post-processing of decoded image and watermarking in wavelet domain of image/video.The achievements of the dissertation include:1. Coding of arbitrarily shaped image is one of core technology in MPEG-4 video compression. We first use low-pass extrapolation to enhance the performance of Katata's approach using classical wavelet transform. In order to make algorithm easier, we study the application of wavelet lifting scheme in transform coding of arbitrarily shaped image objects. We present three new approaches based on lifting algorithm. One needs extrapolation of the image object. The second is a shape-adopted algorithm by modifying the compute of coefficients near the object edge. The third is a shape-adopted algorithm based on modifying boundary extension method of the lifting scheme. The experiment result shows that the image compression performances of approaches using lifting scheme, especially the third one, are better than Katata's approach using classical wavelet transform and the approach using SA-DCT.2. In this dissertation, the motion estimation is combined with 3-D wavelet transform for video coding. In classical 3-D wavelet transform of video, wavelet transform of time domain is applied to pixels in the same spatial coordinate of each frame. But in this paper wavelet transform is applied to pixels in one motion trajectory which is constructed by motion vectors in order to fully utilize the redundancy in the time domain. Because the length of motion trajectory is arbitrarily, a new wavelet transform method for arbitrary length signals is needed. A length-adopted wavelet transform algorithm based on modifying boundary extension method of the lifting scheme is presented in the paper. The experiment result showsthat for video sequence with wide range of movement the compression performance of this approach is better than approach using classical 3-D wavelet transform.3. In this dissertation, we use de-noising by integer wavelet thresholding in post-processing of decompressed images. In Donoho's de-noising algorithm, the high/high portion of wavelet coefficients can be used to estimate noise level and threshold value. If the image signal has abundant high-frequency details in high/high portion of wavelet domain, the estimation of threshold value will be interfered. So we use adaptive transform derived from lifting scheme to make distribution of signal energy more concentrative, so that the threshold value can be estimate more efficiently. We also incorporate adaptivity into redundant transforms to improve the de-noising performance. From the experiment results, we can see that the PSNR values of recover images after DCT and wavelet transform coding are increased. This algorithm can efficiently remove blocking effect.4. We present a new strategy to embed watermarks t...