Digital processing and communication technologies make it possible to realize Video on Demand (VOD) systems. This paper describes the stream control method for VOD system. At first, synchronization method between PCR packets and stored MPEG stream to reduce jitter of output stream is considered. Video synchronization method is also considered between the server and the terminal using NPT (Normal Play Time) standardized in DSM-CC. Furthermore, output method by common format for TS over ATM, standardized in DAVIC, and PS over IP which may use for internet is considered. Finally, video store format and stream control method using NPT is proposed for solving these problem.

View full abstract

A clustered video server, which consists of a group of disk control parts and I/O control parts connected by a network, can get higher I/O bandwidth of a file. This paper describes a multiple-access control method of clustered video server using ATM switch. We present the proposed multiple-access control method can provide following features. (1) Maximum delay time of reading media segment is guaranteed. (2) Multi-channels with arbitrary transmission rates are able to be supported. (3) Priority access and non-Priority access are able to be used at the same time.

View full abstract

A Digital Video On Demand system has been developed in CRL Seika communications research center. MPEG2 is used for video compression in this system. In this paper, we report some result of experimentation about quality of VOD pictures evaluated by using DSCQS methods.

View full abstract

We propose an multimedia application system named the VAmail system (Video & Audio mail system) which combined e-mail system and VoD service to provide high-quality moving pictures. We describe the concept of VAmail system and system configuration. And It is necessary for world-wide service that VoD system should be networking with each other. We discuss about problems for this purpose and a solution of networking VoD systems.

View full abstract

The graphical human-computer communication needs a highly efficient encoding image method because of a large amount of information. For example, JPEG, MPEG algorithm and Fractal Image Compression (FIC) [2] that uses a locally self-similarity in the image are exist. We know that the fidelity of decoding image using FIC is influenced by encoding process for the edge-parts in the image. As before, we discuss some improvements for FIC to compute the collection of optimized transformation for the range block [5]. In this paper, we suggest an application of the adaptive grid generation for FIC to optimized encode the edge-parts in the image.

View full abstract

In the recently information-oriented society, the information of images becomes more important. The information of images has a large amount of information. In the point of the memory size and the communication speed, we need an efficient image-encoding method. Now, the JPEG algorithm is mainstream of image-encoding method. Fractal Image Compression (FIC) that uses a locally self-similarity in the image has been developed as substitution for the JPEG algorithm. In this paper, we suggest the parallelization of decoding process of FIC using the Chaos Game Algorithm.

View full abstract

The existing exact image reconstruction methods for cone-beam tomography compute the 3-D Radon data from the measured projections and then invert it to obtain a reconstructed image. However, the explicit computation of the 3-D Radon data causes large discretization errors and forces intensive computational requirements. We propose a new exact reconstruction method which does not require the explicit computation of the 3-D Radon data. The new method consists of the following two steps. First, we successfully divide the orbit into multiple suborbits and apply the Feldkamp-type filtered backprojection method to each sub-orbit. This step generates multiple incomplete images. Second, a final reconstructed image is synthesized in 3-D Fourier space from the multiple images based on the projection slice theorem. This method is advantageous over other methods in terms of discretization errors, computational requirements, and simplicity of implementation.

View full abstract

Image reconstruction from wavelet extrema representation is useful to develop various edge-based image processing algorithms as shown by Mallat et al., The previous works used the convex projection method for this purpose. We propose a new image reconstruction method which is based on solving a linear equation relating the multi-scale wavelet transform (or the original image) and the wavelet extrema representation. We confirmed that the new method provides better images than the convex projection method by simulation studies. Furthemore, we use the standard regularization theory to suppress the amplification of noise contaminating the wavelet extrema representation. Some experiments of image reconstruction and application to image coding demonstrate the validity of the new method.

View full abstract

Approach of digital image segmentation using network flow in network theory was constructed by Gomory and Hu. The method is to divide the graph some sub-graphs, solve each one hierarchically to apply the multi-terminal maximal flow problem. Consequently that is much faster than non-hierarchical algorithm. In this paper, we propose the method that makes a graph smaller by using improved simple minimum cut algorithm. Using this before that Gomory-Hu algorithm, we can take less time to process image segmentation.

View full abstract

A new method for individual identification from face photographs has been described. The face neuron based identification method of individuals utilizes physiological knowledge on face neurons of an ape which respond to a human face. Twelve major face neurons are represented by face features which are expressed with the distance between two face parts such as a mouth, nose, eyes, contour of the face and so on. Identification of individuals by a neural network is also attempted, although both methods make use of the same face features which are extracted from a photograph by a computer. Thirty four subjects have been identified their faces correctly according to the face neuron method. Even when a person whose hair was intentionally changed was a subject, he was identified correctly by introducing the second identification procedure in which face features relating"hair"were excluded. On the other hand, 85.29% correctness has been obtained, when the same face features are given to the input layer of the neural network which is composed of three layers. In conclusion, the proposed face neuron method is useful for identifying individuals from a photograph.

View full abstract

A method of compressing multi-level images using binary arithmetic coding is proposed. Arithmetic coding of digital images applies usually Markov model to modelize the source image and eliminates the correlation between image pixels. The conventional method uses a fixed state of Markov model for every bit of a pixel. In our proposed method, good compression ratio is achieved by using different state of Markov model for each bit of a pixel, and by introducing the MDL criterion, furthermore improvement of the compression ratio is achieved.

View full abstract

First of all, two evaluation methods are compared with each other. One deals with the way in which a pair of test sequences are presented simultaneously on the same type of two HDTV monitors and the other the way they are presented successively on a single monitor. New viewing conditions modifying the conventional ones for a single monitor are proposed. Secondly, sensitivity to impairments in the test using two monitors is compared with that using a single monitor, and it is found that the method using two monitors has both merits and drawbacks. A new method to improve it is proposed and validity of the new one is verified.

View full abstract

We have already proposed CEFLAR : Congestion Estimation Feedback by Limited Acceleration-Rate/-Ratio. This paper clarifies the performance of CEFLAR on transition characteristics in single-node congestion, fairness index and network throughput in multi-node congestion by simulation. And this paper compares CEFLAR with ER (Explicit Rate) method specified in the ATM Forum. As a result, this paper shows that CEFLAR is inferior to ER method in transition characteristics and fairness but CEFLAR is superior to ER in network throughput.

View full abstract

Recently developing OAN (optical access network) architecture, an optical access system is going to be introduced to the access networks. The most significant requirement is to obtain cost effective solutions to spread the OAN over the access networks. A PON (passive optical network) is one of the attractive solutions. Our objective is to construct a platform equipment for the future OANs considering the trend of the OANs. We have examined the ATM-PON based access network architecture that can be flexibly adapted to the various access networks and developed an OAN system based on that architecture. In this paper we describe the architecture and the prototype of the proposed system.

View full abstract