中口 俊哉

本研究はVirtual Realityを用いた手術トレーニングシステムの開発に求められる臓器のリアルタイムな変形モデルを提案する.提案モデルでは構造力学に基づき臓器内部に梁を配置し,その梁を中心軸とする座標空間を構築する.梁は外力によりひずみを生じ,梁のひずみに基づき座標空間全体のひずみを得る.求めた座標空間のひずみに従い臓器を構成する頂点を移動することで臓器の大域的変形を再現することができる.提案モデルでは臓器が体内に対し固定点を持つとし,その固定点を座標空間の原点とする.中心軸となる梁は原点より臓器の3軸に基づき配置する方法,原点と外力の作用点を結ぶ直線上に配置する方法の2つを提案する.また臓器の応力の減衰を考慮し,梁のひずみにより生じる座標空間のひずみを減衰させることでリアルな臓器の変形を再現する.提案モデルは座標空間のひずみを用いることで,臓器を弾性要素に分割する必要がないため従来の変形モデルに比べ少ない計算量で実行可能である.本論文では腹腔鏡下手術トレーニングシステムに提案モデルを実装し,リアルタイムな変形を確認した(著者抄録)

We proposed a method to simulate the skin spectral reflectance based on an empirical model. The model was constructed based on the simulated data by Monte Carlo simulation of light transport in the skin model.

In the cosmetics industry, skin color is very important because skin color gives a direct impression of the face. In particular, many people suffer from melanin pigmentation such as liver spots and freckles. However, it is very difficult to evaluate melanin pigmentation using conventional colorimetric values because these values contain information on various skin chromophores simultaneously. Therefore, it is necessary to extract information of the chromophore of individual skins independently as density information. The isolation of the melanin component image based on independent component analysis (ICA) from a single skin image was reported in 2003. However, this technique has not developed a quantification method for melanin pigmentation. Ibis paper introduces a quantification method based on the ICA of a skin color image to isolate melanin pigmentation. The image acquisition system we used consists of commercially available equipment such as digital cameras and lighting sources with polarized light. The images taken were analyzed using ICA to extract the melanin component images, and Laplacian of Gaussian (LOG) filter was applied to extract the pigmented area. As a result, for skin images including those showing melanin pigmentation and acne, the method worked well. Finally, the total amount of extracted area had a strong correspondence to the subjective rating values for the appearance of pigmentation. Further analysis is needed to recognize the appearance of pigmentation concerning the size of the pigmented area and its spatial gradation.

In the present paper, we propose a technique to control the appearance on an object in real time using a high-luminance PC projector and graphics hardware. We have previously proposed an image projection technique to reproduce the appearance of a real object on a mock object using a high-luminance projector. By controlling the projected image, the reflected radiance on the mock object is matched to that on the real object. However, in our previous study, only preliminary experiments were performed by matching the appearance empirically. The present paper uses ray tracing to reproduce the distribution of the reflected radiance for appearance matching. It is necessary to measure the bi-directional reflectance distribution function of the objects and the geometry between the projector and the observer's eyes. Since the observer's eyes move with the head in evaluating the appearance of the object, we performed a real-time reproduction of gloss appearance with the movement of the observer's position. The observer's position is detected by an electromagnetic position sensor. Graphics hardware is used in the real-time reproduction to calculate the ray tracing at high speed and render the appearance according to the observer's eye position in real time. Observer rating revealed that little difference in appearance was perceived between the real object and the projected mock object.

A simple 3D virtual mirror interface system is proposed for acquisition and reconstruction of facial pattern, which enables a real time tracking of 3D facial pattern. The movement of lip in the face, which can be estimated based on mean shift and Kalman filter is introduced to recognize the rotation and parallel shift of face. Modified Photometric Stereo Method (PSM) is also proposed to acquire the 3D shape without the influence of deficiency in shadow area of the object. It is shown that the proposed system is effective to acquire and reconstruct for the real-time tracking of 3D facial pattern image than traditional approaches.

We developed a multi-spectral scanner by using LEDs array with different spectral radiant distribution to measure high accuracy spectral characteristics of printing proof. Five kinds of LEDs were selected from the combination of 40 LEDs on the market to minimize the color difference Delta E*(94) between measured and estimated reflectance spectra of 81 kinds of color charts by using polynomial regression and clustering methods. Reflectance spectra of 928 color charts were measured and estimated by using the scanner and the Wiener estimation method. As a result, average color difference Delta E*(94) was 1.23 when 81 color data were used to calculate the Wiener estimation matrix.

In this paper, we propose practical algorithms for the pigment color separation and synthesis of the human skin color to be used on mobile phone with camera. These algorithms overcome the limitation of the conventional algorithm such as using polarizing filters. The pigment colors are separated based on the proposed empirical pigment vector. The empirical vectors are extracted from skin color database. We evaluated the proposed algorithms for 45 images (5 persons, 3 cameras, 3 illuminants). The evaluated results showed that the proposed algorithms performed well on most of skin images.

The goal of this research is to construct the X-ray image output system using ink-jet printer. In order to apply the ink-jet image to medical imaging such as X-ray images, the sufficient quality of ink-jet images is required. In the first step, we measure the tone reproduction characteristics, Contrast Transfer Function (CTF), and Wiener Spectrum (WS) of ink-jet image and conventional silver halide film. The second step is the subjective evaluation performed by medical doctors using medical images printed on silver halide films and ink-jet images. We evaluate the reliability of ink-jet output system based on the results from the receiver operating characteristic (ROC) analysis.

We propose a new model to predict the color reproduction of digital halftone image based on the physical model of dot gain. In the previous papers, we reported that the transparency image of halftone is not influenced by optical dot gain. On the basis of this experimental result, in this paper we analyze the optical and mechanical dot gain separately by using the optical microscopes which can take the transparency and reflectance images of the same area. Transparency images of an ink dot are taken with an optical microscope with a six-band camera and the spectral transmittance of each pixel in an ink dot is estimated by the multiple regression estimation method. This obtained spectral transmittance is converted into the amount of cyan, magenta, and yellow (c,m,y) inks in each pixel. Then we can estimate the shape of ink dot by polynomial fitting of ink amount. The transmittance of c,m,y inks of printed images is estimated by using the proposed method and compared with that of practical printed images. This results show that the proposed method is significant to predict the density of inkjet images.

Hysteresis neural networks are one of the effective heuristic algorithms for constraint satisfaction problems. To overcome a serious defect of HNN which is called a periodic solution, several algorithms have been proposed. This paper describes a comparison between two algorithms of them based on tabu search. One is a previously proposed algorithm named dynamic time constant tabu hysteresis neural networks. Another is a novel algorithm named dynamic equilibrium point tabu hysteresis neural networks. These algorithms are estimated from their performances and implementation costs.

Hysteresis cellular neural networks are one of artificial neural networks which work effectively against large scale problems. In the previous work, remarkable methods have never been developed to overcome the defects of hysteresis cellular neural networks. We then propose a novel architecture for combinatorial optimization problems to overcome them. Experimental results indicate the efficiency of the architecture.