In vector quantization(VQ) of image at low bit rate, edge degradations are observed. To preserve the edge integrity, VQ with classifier has been proposed, which is known as CVQ(Classified Vector Quantization). In CVQ, the construction of classifier is difficult and important.
Wavelet transformed image is divided into sevaral subimages according to direction of the edge and resolution of the image. Therefore, image coding using CVQ in WT domain provides good performance with even simple classifier.
Because pixel values in WT domain have generalized Gaussian distribution, in order to achieve good image quality at lower bit rates, coefficients with the value less than threshold can be removed for each block. And then, the image is coded using CVQ. Until now, vectors are formed from image by the square shaped block. In this thesis, we propose vector formation method in accordance with the direction of the edge. Also, the correlation between other resolution is utilized in coding using CVQ.
