This paper presents techniques for the synthesis and real time animation of tight fitting apparel on human figures within a virtual space teleconferencing system. Physical modeling and animation techniques for cloth, though effective, are too complex to achieve real time results. The emphasis in teleconferencing is more on real time visual realism than physical accuracy. In our approach clothes are modeled as textures mapped onto a geometric skin. Characteristic wrinkled textures for different garments are synthesized and registered on the skin in a semiautomatic and interactive fashion. Wrinkle formation is carried out by texture morphing controlled by the posture of the animated human. Enhanced realism is illustrated within a teleconferencing system by employing hardware pixel blending to morph between synthesized cloth textures.

View full abstract

In computer graphics, it is often required to modify the 3D objects into variable shapes based on the designer's indications This paper describes the modification method of 3D objects defined by 3D polygons in the computer. The main features of the proposed method are summarized as follows. 1. Both partial and entire modification of 3D objects into any shapes can be attained. 2. Modified shapes are controlled by simple cosine functions. 3. Continuous modification can be attained and it generates well balanced modified shapes. 4. Modification can be executed successively by the automatic polygon partitioning. 5. It can be applied to 3D montage systems and so on.

View full abstract

There are some typical methods for reduction of processing time of morphology operation, but these methods have some restrictions on the shape of filter or are not effective if filters have complicated shapes. In this paper, we decompose image and filter by gray scale value in advance for imposing no restriction on the shape of filter and for reduction of processing tine. As a result, this method is robust in filter size and shape, and is effective for complicated-shaped filter which has gray scale value.

View full abstract

Motion estimation is one of the important technology for signal processing in moving picture coding. "Appearance of parallel motion" which is inspected by block matching method contains information of rotation, expansion and contraction as well. Authors propose an estimation method basing on Helmert's transformation with using the "appearance of parallel motion". This paper explains both logical and experimental review of estimation precision for the "appearance of parallel motion".

View full abstract

In this paper in propose a, method for accelerating the fractal block-coding of image. The method reserve a mean-value of pixel-value for each 4 × 4 block of given image f, and construct approximate image f as a pre-processing. The proposed method searches similar a, region for earch range-block by using (1) constant scale-value, (2) f and (3) lower bound of square-error. Computer expriments show that this method drastically recluses the computation time for the fractal block-coding.

View full abstract

This paper showes a neuromagnetic source analysis by using 64 channel SQUID system and MRl information. The magnetic source is modeled as current dipoles, and the head is modeled as 21 conductive sphere. These models make it possible to calculate the magnetic field produced by the impressed current and the volume current. The number of dipoles is estimated by MDL criterion. The least squares cost function to specify the dipole parameters is minimized. After computer simulations. Auditory evoked magnetic sources are localized by utilizing the anatomical MR image. Estimated dipoles are combined with the 3-D MR image to visualize and understand the neural activity.

View full abstract

We describe a prototype system that provides the anatomical information of blood vessels using cone-beam CT. Digital angiograms are collected in a short period of time and three-dimensional (3D) image of blood vessels is approximately reconstructed by a short scan cone-beam filtered backprojection algorithm. The analysis method of the 3D blood vessels image is based on a graph description of 3D blood vessels images using a thinning algorithm for 3D binary image and a distance transformation of 3D line pattern. From results of the application to patient's abdominal blood vessels, we present the effectiveness of the system.

View full abstract

Lung cancer is known as one of the most difficult cancer to cure, and the number of deaths that it causes is usually increasing Recently, helical CT images are used for the mass screening process as a diagnosis tool, but considering the number of images to be checked, it takes long time to check those images. For that reason, we develop a computer assisted diagnosis system that detects the lung cancer, candidates in its early stage from helical CT images with specified measurement conditions, which allows to make wide mesurements in a short time. Our diagnostic algorithm is based on image processing techniques and medical knowledges. From results of the application to the patients of lung cancer, we present the effectiveness of our algorithm.

View full abstract