A new picture coding method (Adaptive Block-size Vector Quantization) is proposed for small screen video communication using 64kbps transmission line. This coding method uses two different size blocks and selects them adaptively depend on distortion between the input vector and codeword. The coding simulation of a facial image is carried out and these results are shown.
The variable block size coding algorithm, which has been proposed by the authors for the encoding of monochrome pictures, is applied to the encoding of color pictures. This coding algorithm is applied to the luminance signal Y and chrominance signals U and V. Here, the luminance signal is used as it is but the resolution of chrominance signals are reduced to 1/4. Computer simulation results for coding efficiency show that this coding algorithm can be achieved about 0.60-1.02 bits/pel for color pictures.
This paper presents an algorithm of stereo matching using relaxation technique based on the relative order relations between projected points and edge region connectivity. The algorithm consists of following four main steps : 1) Segmentaion of the images into the connected edge regions, 2) Calculation of the initial correspondence probability within the scanning lines, 3) The correspondence probability is renewed relaxation technique using the region line up uniformity and the edge region connectivity, and finally 4) Determination of the correspondence between egde regions in each scanning line and extraction of the three dimentional information. Experimental results on this stereo matching method show that the matching accuracy is 72.34% to 94.07% for the different type scenes.