When viewers look at stereoscopic display images, they do not feel very natural. The factors of unnatural feeling are considered to be the distortion of images and fusion area. In this study, previous factors under various image input-output conditons are estimated using computer simulation, and the relation between the results of the simulation and the unnatural feelings are discussed.

View full abstract

Sub-band image coding is Considered to be a promising technique for high fidelity compression of high resolution images, in the traditional sub-band image coding algorithm where a scalor quantizer is used, however high frequency sub-band signals are usually discarded. Which Incurs rather noticeable degradation of reconstruction quality. To solve the above problem, the work herein introduces improved Permutation codes, recently developed by the authors extending the concept of prototypal permutation codes, into sub-band image coding, and applies improved permutation codes to quantization of high frequency sub-band signals. The simulation results demonstrate that improved permutation codes are more useful than an entropy-constrained scalor quantizer as a means for quantizing high frequency sub-band signals, Moreover, the work herein develops an adaptive quantization technique combining improved permutation codes and entropy-constrained scalor quantizers, and demonstrates the usefulness by computer simulations.

View full abstract

This paper describes an image coding method combined with moving object detection, motion compensation by mapping, and orthogonal transform for arbitrary shapes. This method does not use grid shape block division, therefore there is no grid shape block distortion. Because uncontinuous parts in the inverse transformed signal by the arbitrary shape transform appear on the edges of the segmented objects, this type of distortion is hardly perceived. Consequently, this coding method can give subjectively good quality reconstructed images.

View full abstract

Although DCT coding has a very good coding performance, it causes block distortion at high compression ratio. A DCT scheme with variable block size is effective to reduce the block distortion. In this paper, we propose a DCT-based coding of the complementary pyramidal image. it applies DCT with fixed block size to the complementary pyramidal image and realizes DCT with variable block size effectively.

View full abstract

This paper presents a concept of the 3-D Markov model coding which uses correlation between bitplanes. The proposed method is evaluated by means of measurement of the Markov model entropy using seven 512×400-pel images with 16 levels. The simulation was performed by using two transmission schemes, which are breadth first and depth first, and two bit-plane representation schemes, which are natural binary code and gray code. Simulation results show 3-D Markov model can reduce the entropy by 10-20 percents compared with the conventional Markov model. Finally, this paper shows that the 3-D Markov model using common reference elements can also effectively reduce the entropy.

View full abstract

This paper describes a variable rate moving image coding system for video distribution services using the asynchronous transfer mode (ATM). The system is based on band division, and permits high performance coding of moving images as well as control of maximum coding rate. It can also deal with cell loss and usage parameter control in ATM networks. In simulation where the system was evaluated using test pictures, it was found that the average coding rate was 20 - 30 Mbit/s, and that the coding rate could be controlled to suppress picture quality deterioration.

View full abstract

An image synthesizing method for the conference system with realistic sensations is described. In this system, a human image in virtual space has to be synthesized by participant's operation and tracing the motion of human in real space. Intention of operation can easily be represented by language. On the other hand by image and sensor, information about the motion of human in real space be detected well. We proposed a new image synthesizing method using an associative memory. the memory memorizes human motion patterns to integrate these information, and to to conpensate abstractiveness of language information and insufficiency of image and sensor information.

View full abstract

This paper presents a new image description method based on the region segmentation. Segmented images are described by DIME-structure and coded by chain-code of each contour, the average value of RGB, coordinates of vertex, and so on. Accoding to the experiment, it is shown that browsing of segmented images gives the efficient image retrieval.

View full abstract