In motion compensation, the quantization stepsize of motion vectors is a critical parameter that determines the properties of the motion compensation method. For example in block matching, the prediction characteristics improve substantially by reducing the quantization stepsize from single pel distance to half pel distance. In this paper, we point out that there are three quantization parameters in generic motion compensation methods. The effect of each parameter on the prediction characteristics of motion compensation is examined by simulation.

View full abstract

A new design method of Motion Compensation (MO filter is introduced. The design is performed for the purpose of reducing blurring called "drift". The drift is occurred in the scalable decoder which decodes a part of the bitstream obtained from MPEG encoder. This paper analyzes the MC in the scalable decoder from multirate signal processing stand point and describes the MC in z transform domain. It also describes a new motion compensation which can reduce the drift. Simulation results show the effectiveness of the proposed method.

View full abstract

It is consiclered that high visual quality of image coding system is realized by adopting human visual system effectively. And then, in this paper, we propose the image conding control scheme with color difference in uniform color space. At first, we consider the use of the color difference value between original image and coding image was reduced by comparison between proposed schemes and the TM method, we confirmed the coding control human visual efficiency in proposed method.

View full abstract

This paper proposes a processable video coder based on camera operations. The coder intends to provide processibility on decoded video data and to improve coding efficiency. These goals are realized by separating the input sequence into some structural elements: camera parameters, moving objects and a background. In this paper, it is assumed that the subjects taken by a camera are stationary on a plane which is parallel to the imaging plane in the camera. The simulation results show that the proposed coder generally gives favourable editing flexibility and significant improvement of the coding efficiency. The results also reveal that the reconstructed image quality becomes around 50 [dB] when the accurate parameters are available, and that it falls significantly as the error of parameters accumulates.

View full abstract

We propose a new error-resilient very-low-bitrate video coding scheme. This scheme has the carefully designed syntax so that it can utilize a self-synchronizing property of the variable length code. By encoding the proposed scheme, the decoder resynchronizes after the error detection and gets the codewords of the following part of the bit stream correctly form the view point of syntax and semantics. The simulation result shows that the SNR gains about 1~4 dB by the proposed scheme compared with by conventional scheme.

View full abstract

One of the most important topics of recent years is use of multimedia technology in radio systems for video, voice and data transmission. On the other hand, a major problem in such systems is the occurrence of errors due to multipath, interference noise or shadowing. So in this paper, it is clarified the effect on moving picture due to signal error occurred in radio channel by computer simulation.

View full abstract

The present status, requirements, system configuration, specifications, etc. of the conditional access system for digital broadcasting are investigated. The mainly considered signal scrambling method, which can be commonly used for various kinds of signals of digital broadcasting, is that the PN sequence is added to the Transport Stream of MPEG-2. The transmission method for the entitlement control messages, which uses the section specified by MPEG, and the items which are necessary for the standardization of the interface between the IRD and the detachable security module are investigated.

View full abstract

Discrete cosine transform (DCT) is very useful as the key-point tool of high efficiency image coding. We have presented a new computational algorithm of the DCT-1 that is led by the same method as the multirate Forurier transform. In this algorithm, each sampling rate for each signal component is used and extreme values of each signal component are sampled and accumulated by adding or substracting. The DCT coefficients can be calculated by dividing the accumulated values by the sample number. This new algorithm needs only one multiplication for one DCT coefficient. In this paper, we introduce a basic principle of this algorithm and another algorithm which does not have compensation terms in calculation. And we extend them into the two-dimensional algorithms and check the accuracy of those algorithms by computer simulation. And more, we consider a hardware realization for this algorithm and then show the possibility of real time processing of this hardware implementation.

View full abstract

In this paper, we propose a rate control method by adaptive quantization table. The quantization table is decided in each frame by edge detection in current frame, total bits and scale factor in privious frame. This method is appropriate for motion picture compression using JPEG chip which has DCT, quantizer and Hoffman coder.

View full abstract

We have studied merits and demerits about two kinds of VOD(Video on Demannd) servers which were classified by a method of how to connect disks stored compressed digital video files and we show features of each architecture. One of them is a distributed disks VOD server. Another is a shared disks VOD server. The distributed disks VOD server consists of title-management unit, several stream delivery units and ATM(Asynchonous Transfer Mode) switch. We have developed a prototype of the shared disks VOD server for verifying the features which are system scalability, various trick play specification and command high-speed response, and making a experiment apparatus for developing a title-management method. We have increased network efficiency and realized interactivity network achitecture. We have developed VOD server with high-speed CPUs, high-speed disks and ATM. As a result, we could provide 50 chanels(1.2Mbps) × N(stream delivery units) at the same time.

View full abstract

A new type of liquid crystal(LC) projector has been developed, utilizing the human vision characteristics :a lower spatial resolution for the colored pattern compared to that for monochrome pattern. The project produces two kinds of signals from the video composite signal or the regular RGB signal, processes and applies them separately to LC panels for the color and for the luminance, and then superimposes the color image and the luminance image optically, restoring the original image. The low frequency color channel, especially color panels, can be of low cost and thus the low cost projector can be realized. The developed system has three color panels for R,G and B and a luminance panel, all of which are of 1.3 inch diagonal size, and two polarizing beam splitter cubes;one is to split light from the lamp into s and P components and the other to synthesize the two components after passing through the color and the luminance panel, respectively. The obtained illumination is 600 lux for the image of 40 inch screen size with a metal halide lamp of 250 W.

View full abstract

In the realization process of the color image data base, extracting characters on the color image is very important aspect in auto-indexing. In this paper we present a method based on the variance in the color space to extract the characters from the color image. In this method, first we transform the image from the RGB color model to the L^*a^*b^* color model, which represents "visiual color difference" as Euclid distance, and observe the distribution density of each color in the color space. Then if we assume that all characters have the same color and the background image is a natural color image, the distribution of the color of the characters is expected to be concentrated in one point in the color space. Using this fact we can estimate the characters' color in order to extract them. However, the pixel values actually show some variation even when the character color seems to be the same. We solve this problem by using a mean filter to average the pixel values. As a result, the proposed method shows good ability for extracting the characters from the color image.

View full abstract

The problem of detecting lines in a digital image is one of the most fundamental problem. in pattern recognition. The Hough transform(HT) is well-known technique for its effectivity to detect straight lines from a noisy image. The standard HT consists of two stages, the first stage is voting and the second one is the peak detection in the line parameter space. The voting is performed in the (ρ,θ) line parameter space with the same weight for all θ, and the space is quantized to accumulate the voting from a digital image. To get more reasonable result of line detection, we introduce the linearity in local region, and propose a method of voting the linearity into parameter space. The experiments show that our voting method gives the robust result of line detection even if a given image has characters as well as lines.

View full abstract

The electrostatic field in an electrophotographic development zone is calculated using a newly developed method. The zone is assumed to be composed of four parallel layers, namely, a conductive substrate, a photoconductive layer, a development space, and a development electrode. The equation to describe the contrast potential was obtained. And numerical analysis and the equation results coincided with each other. It is clearified that the parameters effective for increasing electrostatic contrast have been the bias voltage V_B, ratio of the relative dielectric constant have been the bias voltage V_B, ratio of the relative dielectric constant ε_<r3> ε_<r2>, photoreceptive film thickness a , and prosecution volt-age V_R.

View full abstract