Research Of Multiwavelet Transformation In Image Coding

Image transmission and storage have a bright future in visual telephone, digitaltelevision, remote medical treatment and satellite remote sensing, but the mass imagedata limits the communication bandwidth and storage media. To adapt thecommunication bandwidth and storage media in use, image data must be compressedby being coded properly and removed redundant information. Most of mature encryption algorithms, e.g. JPEG, H263, MPEG, take DCT astheir core, which brings in geometric distortion and has "block" effect with highcompression rate. Wavelet transformation is newly used in signal analysis, and zerotree codeguarantees its success in image compression. Compared with DCT, image codingbased on wavelet transformation has higher compression rate and signal-to-noise ratiowithout "block" effect in low code rate as in JPEG reconstruction image. It hasbecome a critical part in some international image compression standard such asMPEG-4. But wavelet code is complicated and has latent image;and when thecompression rate is extremely high, the reconstructed image may have distortion. One of the key researches in wavelet code is to select the right wavelet base. It isimpossible for single wavelet to have orthogonality, symmetry, short support and highorder of vanishing moment at the same time, which are requested by image and visualcharacteristic for human being. Theoretically, multi-wavelet can own these featuresmentioned above simultaneously, which makes it the main research direction. This paper introduces multi-wavelet theory, structure, transform and encryptionalgorithms, and also explores the design of the pre-filter. The main contents include:1,Exploring the pre/post-signal processing as theory in multi-wavelet analysis, andputting the pre-processing methods forward. Introducing Good MultifilterProperties(GMPs) in the multi-wavelet filter design for image compression.2,Bringing forward a vector quantization multi-wavelet zerotree code on the basisof analyzing single wavelet zerotree code and considering the multi-waveletcharacteristic. Taking the multi-wavelet analyzing of image as vector quarter treestructure because of the correlation among sub-blocks after multi-wavelettransformation. Rearranging the multi-wavelet coefficient to maintain the zerotreecharacteristic. The image compression simulation justifies the better quality ofmulti-wavelet than original wavelet does.