We propose a systematical color image coding which will be very important is near furture image communication via ISDN. The "systematic" means systematic design and signal processing based on the Picture Quality Scale(PQS). PQS is defined as a function of the color difference which is expected to be defined as an Euclidean distance of signal error between the original image signal and the decoded signal. We should start the following detailed discussion as soon as possible; 1. Quantization of RGB signals and HVC signals under the condition of "quantization error is not perceptible". 2. (a) Discussion on uniformity of Munsell Renotation System and modification if necessary. (b) Discussion on the precision of the mathematical two-way transformation between RGB color space and HVC color sace. (c) The development of the two-way transformation between RGB color space and YCMK color space.
A new type of hybrid subband coding method using motion compensation (MC) is described in this paper. Energy compaction of lapped transforms and analysis/synthesis filter banks such as QMF are higher than that of Discrete Cosine Transform. However, when we combine the intraframe/MC-interframe adaptive prediction and filter banks, the problems of 1) level gaps between two successive MC blocks (caused by block matching process), and 2) level gaps between intraframe and interframe blocks should be solved. Because those level gaps contain very high frequency components, they reduce the efficiency of subband codings. Thus, we propose a hybrid coding structure, where motion estimation is carried out in the image domain, and interframe differences are taken in the frequency domain. In this method, interframe prediction and intra/inter switching can be carried out in the frequency domain. Thus, level gap problem does not appear. In addition, we show that the process of loop filtering can be carried out by only discarding high frequency band data of prediction images. By using this method, the coding efficiency of proposed MC+subband coding can be enhanced. Also, an adaptive loop filtering can be performed by varying the number of bands to be used for interframe prediction.
A new type of motion compensated subband coding method suitable for wavelet transform and octave spaced QMF bank is described in this paper. Although the energy compaction of wavelet transform is comparable to Discrete Cosine Transform, wavelet transform does not produce artifacts such as blocking effects and mosquito noises, which are most typical problems in the DCT coding. However, when we combine the intraframe/MC-interframe adaptive prediction and wavelet transform, the problems of 1) level gaps between two successive MC blocks (caused by block matching process), and 2) level gaps between intraframe and interframe blocks should be solved. Because those level gaps contain very high frequency components, they reduce the efficiency of subband codings. Thus, we propose a hybrid coding structure, where motion estimation is carried out in the image domain, and interframe differences are taken in the frequency domain. In this method, interframe prediction and intra/inter switching can be carried out in the frequency domain. Thus, level gap problem does not appear. In addition, we show that the process of loop filtering can be carried out by only discarding high frequency band data of prediction images. Also, an adaptive loop filtering can be performed by varying the number of bands to be used for interframe prediction.
A variable bit-rate HDTV coding algorithm based on motion-adaptive DCT is investigated for ATM environments in B-ISDN. Adaptive 2-layered coding, an ATM cell matrix for error correction, and a block interleave for error concealment are proposed to keep picture quality high by compensating for ATM cell loss. A new VLC and a congestion control scheme that restricts peak-rate and average-rate are also proposed for traffic control. The proposed algorithm is shown to reduce the coding bit-rate for HDTV conference applications in ATM environments to 10-30 Mb/s.
In the case of VSB-AM cable transmission of the MUSE signal via satellites, three methods are available according to the mode of the modulated MUSE signal : FM mode, AM mode and de-emphasized FM mode. We have carried out subjective assessments using experts as assessors to examine the effects of both satellites channel and cable channel noise. The results show that there is no significant difference among the three methods and the required C to N ratios are higher than those of existing cable facilities. The solutions to this problem are also discussed.
In this paper, we present the automatic extracting method of the object segmentation for moving image. The method is composed of the following two steps : (1) in the first frame, the simple boundary line of the object is given by a user, then the object is detected ; (2) from the second frame, the object is automatically extracted with the result of the preceding frame. The appropriate processes of correlation for matching are discussed and the recognition result is illustrated by a number of examples.
This paper presents a new recognition method for machine parts composed of polygonal and cylindrical faces by using the selective two-stage neural network. An appropriate operation which depends on characteristics of the face is automatically selected and extracts geometric features of the face effeciently in order to recognize the object. The first stage neural network is for identifying object faces and has p-type Fourier Descriptors, area and perimeter as input parameters. The second one is for recognizing the object by a set of aspect graphs which represents the connection among the faces. Consequently, by applying this method to recognize machine parts, the efficiency was improved as compared with ordinary methods.
Recently in the field of computer vision, there have been many attempts to obtain a symbolic shape description of a 3D object by fitting simple primitives to the range data. Since many of these researches required to solve nonlinear minimization problems, they cost so much time and it was difficult to obtain a unique shape description for the given range data. This paper proposes a method to obtain a unique shape description of an object by using wavelet transforms. Wavelet transform is a new technique for signal analysis which decompose a signal on locally distributed orthonormal basis functions. We first decompose a volume data on 3D wavelet basises and then approximate the shape with a relatively small number of functions by using only significant basis functions. We have successfully applied this method to a human head volume data and typical results are shown.
Visually evoked potentials (VEPs) have been elicited by randomdot stimulation binocularly. The 1st randomdot pattern is fully correlative binocularly. The 2nd has partially non-correlative area with each other. The 3rd pattern is that with disparity, i.e. stereogram. Potentials evoked by those patterns have N170 and P240 components. In the case of viewing stereogram, VEP have been increased according to the size of disparity area. The size of non-correlative area of pattern without disparity has no influenced the amplitude. These results have been corresponding to subjective stereognosis.
We measured the fusional time which is the interval from displaying disparity pattern to getting fusion and stereognosis. Two types of stereoscopic pattern were displayed on the CRT monitor. Wire-frame pattern was fused easily within ±2 degree in disparity. Disparity in random-dot pattern would increase the fusional time remarkably. We have proposed a model to explain the fusional time difference between wire-frame and random-dot stereogram.
Featuring registration-free, high-quality pictures and stability the first 3-ccd solid-state high-definition camera HSC-100,WHICH adopts a layer CCD with 2-million-pixels, promises a 6dB S/N ratio and a 6dB sensitibity, far superior to existing TV cameras.