This paper deals with a singular value decomposition (SVD) image coding scheme using Gain/Shape vector quantizers. We named this scheme "SVD-Gain/Shape vector quantization (SVD-G/SVQ)". Previously, we present the optimum structure of the codebook for SVD-G/SVQ, and its design algorithms. In this paper, we present experimental simulations of SVD-G/SVQ. Experimental simulations show that at about 0.5 (bit/pel) SVD-G/SVQ is similar in performance to the adaptive discrete cosine transform image coding scheme proposed by Chen and Smith.
Formerly, we proposed the discrete cosine transform image coding using gain/shape vector quantizers (DCT-G/SVQ) . This paper deals with adaptive DCT-G/SVQ using the adaptive technique by activity index. Experiments show the adaptive DCT-G/SVQ yields a performance advantage of 1-2 (dB) over the nonadaptive DCT-G/SVQ. Beside, the adaptive DCT-G/SVQ yields performance advantage of 0.5 1.5 (dB) over the adaptive DCT coding using both Max-type scalor quantizers and the adaptive technique by activity index.
An Orthogonal Transform Coding (OTC) is discussed, considering the statistics and visual properties in detail. In the design of the quantizer, a perceptual property for luminance and a spatial frequency characteristics of Human Visual System (HVS) are taken into account. New measures to evaluate the performance of OTC are adopted. They are mean square error weighted by an estimation curve of random noise, changes of errors and autocorrelation of errors weighted by spatial freqency charecteristice of HVS. By the aid of computer simulations, the optimum quantizer is designed. This method is especially avariable for 1.0 bit/pixel or more.
Papers in the field of color vision mechanism is reviewed. A recently proposed color vision model which is based on the stage theory and includes nonlinear paths between luminance and two chromatic channels is explained in connection with the luminous efficiency, cone response functions, opponent color process and nonlinear nature in color perception. Finally, Land's retinex theory is briefly described. Accoding to this theory, we can interprete the constancy in color perception and the principle of the Land's two-color method for obtaining full color on a screen by superimposing two images of black and white slides which are taken through either red or green filter and are projected through a red or no filter, respectively.
This paper presents a fast image enhancement method for handwritten drawings. When drawings with little contrast between black and white grey levels are digitized, such troubles as cuts and connections among lines will be caused. Thus, the method is to be proposed to resolve these problems. The method will enhance peaks and valleys of grey levels ditribution among pixels without increasing grey level of noises. Experimental examples show that this method may be of practical use.
There have been various multifunctional workstations developed for office workers or engineers. Most of them can display multiple windows to provide the sophisticated user operations. The display control unit (DCU) with the multi-window has been developed. The DCU has the memory, called "window buffer, " which keep raster image of windows. In this paper window-buffer's functions are discussed, and hardware structure of the window buffer is shown.