An analog method for image restoration is applied to range images. The method searches an image minimizing an energy function of distance variables and line processes. With the aid of continuation method, artifact edges induced by noises can be removed and as a byproduct a multiscale representation of edges can be obtained. Doubling the number of line processes enables us to distinguish convex creases from concave ones.

View full abstract

For calculating curvature of curves, there are three methods, i.e. ones by tangential angle, by differentiation of coordinate and by radius of curvature. Although these methods are equivalent and lead same results analytically, they do not always give same results for digital curves. In this report, we concretely show how they are different each other for digital curves by theoretical and experimental researches.

View full abstract

Boundaries or contours of images include much important information to recognize images. Boundary extraction and description are, therefore, natural and fundamental processing to grasp the properties and characteristics of images. Dominant points are the high curvature points along a digital curve that have important shape features. Many algorithms have been proposed for detecting them. Almost proposed methods, however, require any parameters to be determined in advance, and are weak in noise. A new method for detecting dominant points is proposed. It does not require any parameters beforehand and it is robust in the presence of noise.

View full abstract

An individual identification system is developed. In this system we unify results of profile curve identification and front view image identification to obtain more successful recognition rate. In profile curve identification process, equilateral polygon approximation and P-type Fourier descriptor are made use of. In front view image identification process, density histogram comparison method is used, which takes shorter calculation time. Both of above results are mixed to obtain final results.

View full abstract

We present a new method for the control of a robot manipulator. The method assumes virtually no prior Information about the robot's kinematic behavior and therefore avoids the costly and time-consuming calibration required by many existing methods. Instead, it relies on continuous sensory feedback, integrating visual data and kinematics into a single mapping, called the perceptual kinematic map (PKM), from the robot's joint space to the camera space. We present results from a real-time, software simulation of a manipulator using a PKM controller.

View full abstract

Computer simulation of fractional Brownian motion is presented. The simulation of fractional Brownian motion has been performed by using the successive random displacement method and the Fourier filtering method. The results obtained from the methods are shown graphically.

View full abstract

Generally, the blind persons acquire character information using the raised type, Braille However, it takes time to learn the Braille and it is not always easy to read it. Our object in this study is to develop a system, which will be able to search the Braille text optically, decode and speak in synthetic speech. Our plan in reading modes is not a passive mode such as an automatic scannig, but an active mode where a user can scan himself the text using a scanner attached to his fingerchip. In this paper, we describe on the method of decoding Braille images taken in using the fiber scope. We tried to reconstruct the line images and to decode those in the line scanning mode which is one of two modes. As results, the character correct rate was about 90%. Moreover, both the scanning speeds and the frame hitting rates in the optional scanning mode were examined.

View full abstract

In the case of the presence of the high absorption coefficient object in the image reconstruction area of the X-ray CT scanner, its reconstruction image will have many artifacts. Carrying out the computer simulation, we confirmed that by reconstructing the image using I-ART(Improved algebraic reconstruct ion technique) proposed in this paper, such artifacts suppressed drastically.

View full abstract

The inference to the light source which is obtained from the reflective light is one of the basic techniques of auto white balance control systems,etc.. The inference is realized by selecting the light source, which is supposed by average color differential of an image, however the inferencial method to the mix light was not examined. We propose a method to obtain the kind of the mixed light sources from the ratio of distance on the color differential space between position of an average color differential of an image and positions of light source, and the method which is added the characteristic spectrum on the color differential plane. So that a good inference is obtained for the mixed light source.

View full abstract