Takashi Watanabe

An extension of the generalized Hough transform to realize affine-invariant two-dimensional (2D) shape detection

We present a method for two-dimensional (2D) shape detection applicable under affine transformation. The problem of affine-invariant shape detection is an important and fundamental research subject in computer vision. Although various methods have been proposed to solve this problem, most of those approaches are not well suited for the following general cases: (1) a shape to be detected is occluded by other overlapping objects, (2) a shape boundary is partially broken because of noise or other factors. We introduce a method to deal with such cases, which extends the generalized Hough transform to be an affine-invariant shape detector. This method, called the affine-GHT, utilizes pairwise parallel tangents and basic properties of an affine transformation to carry the direct computation for six parameters of an affine transformation. Experimental results demonstrate that the proposed method performs successfully and efficiently.

Improved dithering methods for color quantized images

A color image produced from a continuously toned color image by using a smaller number of colors (e.g., 256) is called a “color quantized image.” Previously, the author and colleagues proposed a high-speed color quantization method which produces an image of almost the same quality as the original. However, if colors in an image change gradually (such as an image produced by computer graphics), false contouring occurs and deteriorates the image quality. To solve the problem, this paper proposes four new methods, three of which are shown by experiments to have excellent dithering effects which overcome the weakness of the conventional error-diffusion method and are practical.

Detection of broken ellipses by the hough transforms and multiresolutional images

This paper discusses the problem of separating and detecting ellipse patterns. The tradiational ellipse detection methods have assumed contour lines to be connected. Thus, they are disadvantageous in that they cannot detect ellipses whose contours are broken considerably due to noise and overlapping. Here, a new ellipse detection algorithm is presented which can cope with images such that many ellipses of arbitrary shapes are mixed and connectivities of contours are missing completely due to noise and that parts of contours are lost due to overlapping. The proposed algorithm is based on the Merlin-Farber method, a kind of Hough transform, and the multiresolutional images and the least-square method are incorporated. Good detection results were obtained in the experiments implemented to demonstrate the effectiveness of the proposed method.

Effects of systemic administration of 2-(4-phenyl-piperidino)-cyclohexanol (vesamicol) and an organophosphate DDVP on the cholinergic system in brain regions of rats.

Vesamicol is known to inhibit the transport of acetylcholine (ACh) into synaptic vesicles in vitro, but much less is known about its effects in the brain in vivo. To assess the effect of vesamicol in vivo, we examined cholinergic parameters, such as the subcellular distribution of ACh, activities of enzymes, uptake of choline, and muscarinic receptor binding in the striatum, hippocampus, and cerebral cortex of rats 30 and 60 min after intraperitoneal injection of vesamicol (3 mg/kg) or of vesamicol in combination with DDVP (5 mg/kg), which was administered 10 min before vasamicol. The levels of cytosolic ACh increased in all regions of the brain after injection of vesamicol, while those of vesicular ACh decreased in all regions except for the striatum. The increase in the levels of extracellular ACh and cytosolic ACh in the striatum induced by DDVP was generally enhanced after injection of vesamicol, Vesamicol did not reduce the level of vesicular ACh when DDVP had been injected previously. Vesamicol did not induce any significant changes in the activities of enzymes, choline uptake, or binding of [6H]quinuclidinyl benzilate to the muscarinic ACh receptors in the three regions. Changes in the cholinergic parameters caused by DDVP were not reversed by the combined administration of DDVP with vesamicol. The present results indicate that vesamicol can inhibit the transport of ACh into synaptic vesicles in the brain tissue in vivo, although it cannot reverse the effects of DDVP that has been injected prior to vesamicol.

A fast algorithm for color image quantization using only 256 colors

In the recent workstation for CAD and CG, a color map display is often used, which can display 256 colors selected from 224 (approx. 16000000) colors using a look-up table. Even in the computer system with such a color display with such a limited number of colors, it is desired that the continuous color image should be processed. In most cases, by determining colors which are important to the image (representative colors), the color image can approximately be displayed without much degrading the image quality. This paper proposes a new method for the approximate display of the color image by 256 colors. In the method, RGB color space is successively partitioned to determine 256 representative colors, so that the quasi-optimum solution is obtained in the sense of the least-mean-square error. By the proposed method, a highly accurate approximate image is obtained with several times faster speed, compared with the traditional methods. The work memory required in the program is only 100 to 200 kBytes, which is less than in the traditional method.

Piecewise-linear approximation of discontinuous contour images based on clustering procedure

This paper discusses the problem of the piecewise-linear approximation of contour images. Almost all the piecewise-linear approximation methods that have been proposed can be applied only for continuous contours. Methods that can work well for discontinuous contours, for instance, have not yet been found. In this paper, a new method is proposed based on clustering as a piecewise-linear approximation method that can even be applied to discontinuous contour images. Specifically, the method uses hard clustering or fuzzy clustering, where the set of edge points in the image is partitioned into line segment groups and parameters of line segments are computed from the groups. Good results were obtained in the evaluation experiment to verify the effectiveness of the proposed method.

Effect of a single roughness element on boundary layer transition over a wedge

An experimental study was made to examine the effect of a tripping wire on boundary layer transition over a wedge of 15°, 30°, or 60° with pressure gradient. Turbulence measurements were carried out in a range from the near-wake of the wire to the fully turbulent region. The results were as follows: (1) The transition Reynolds number Ret ( = U∞xt/v) atthe most effective condition is given by the formula Ret = 10 Rek + (1.33 × 104), where Rek ( = U∞xk/v) is the Reynolds number at the wire position, U∞ the free-stream velocity, v the kinematic viscosity of the fluid, and xk and xt the location of the wire and the transition point measured from the apex of the wedge, respectively. (2) The minimum transition Reynolds number Retm is a function of the ratio of the height of the roughness element k to the displacement thickness δ∗ of the boundary layer at the location of the element. The magnitude of Retm decreases with increasing k/δ∗ and increases with the increasing wedge angle at a fixed value of k/δ∗.

Compound decision theory and adaptive classification for multispectral image data

Previously, we reported on the experimental evaluation of multispectral image classifiers utilizing local spatial information. It was shown that the two classifiers of compound decision and adaptive classification exhibit a good performance in terms of the discrimination rate and the processing time. This paper discusses the improvements of those two methods further to improve the discrimination rate, and presents the result of experiment to indicate the effectiveness. More precisely, the following improvements were made. For the compound decision, the improved method estimates the context distribution based on the dependency of three pixels. For the adaptive classification, an extended adaptive classification is newly proposed where the adaptibility is improved by extending the region for classification to shapes other than a square. Among those, the extended adaptive classification is shown to be a highly practical method, wherein the discrimination rate is much higher than in the pixelwise classification such as Bayes decision, and it is expected to be employed widely as the standard classification method.