中口 俊哉

In this article, we propose an efficient and accurate compressive sensing-based method for estimating the light transport characteristics of real world scenes. Although compressive sensing allows efficient estimation of a highdimension signal with a sparse or near-to-sparse representation from a small number of samples, the computational cost of the compressive sensing in estimating the light transport characteristics is relatively high. Moreover, these methods require a relatively smaller number of images compared with other techniques although they still need 500- 1000 images to estimate an accurate light transport matrix. Our proposed method - precomputed ROMP (Regularized Orthogonal Matching Pursuit) - improves the performance of the compressive sensing by providing an appropriate initial state, which allows us to more accurately estimate the matrix with fewer images. This improvement was achieved through two steps: 1) pseudo-single pixel projection by multi-line projection - measuring coarse light transport characteristics to utilize them as an initial state, 2) ROMP with initial signal - refining coarse light transport characteristics with the compressive sensing theory with the initial signal. Precomputed ROMP was carried out by parallel processing. With these two steps, we were able to estimate the light transport characteristics more accurately, much faster, and with a lesser number of images. © 2010 IEEE.

In this article, we propose an effective color-processing algorithm to analyze the hyperspectral image of the tongue and its application to preventive medicine by the concept of Japanese traditional herbal medicine (Kampo medicine). Kampo medicine contains a number of concepts useful for preventive medicine such as "Mibyou" - disease-oriented state - signs of abnormalities. Hyperspectral images of the tongue were taken with the system with an integrating sphere, and tongue area without coating was eliminated automatically. Then, spectral information of the tongue area without coating was analyzed by principal component analysis, and the component vector best representing the clinical symptom was found by rotating the vector on a plane spanned by two arbitrary principal component vectors.

The image segmentation is the first and essential process in many medical applications. This process is traditionally performed by radiologists or medical specialists to manually trace the objects on each image. In almost all of these applications, the medical specialists have to access a large number of images which is a tedious and a time consuming process. On the other hand, the automatic segmentation is still challenging because of low image contrast and ill-defined boundaries. In this work, we propose a fully automated medical image segmentation framework. In this framework, the segmentation process is constrained by two prior models; a shape prior model and a texture prior model. The shape prior model is constructed from a set of manually segmented images using the principle component analysis (PCA) while the wavelet packet decomposition is utilized to extract the texture features. The fisher linear discriminate algorithm is employed to build the texture prior model from the set of texture features and to perform a preliminary segmentation. Furthermore, the particle swarm optimization algorithm (PSO) is used to refine the preliminary segmentation according to the shape prior model. In this work, we tested the proposed technique for the segmentation of the liver from abdominal CT scans and the obtained results show the efficiency of the proposed technique to accurately delineate the desired objects.

In the present paper, we have studied the relationship between ambient illuminations and psychological factors in viewing video image. Four kinds of video images with changing layouts of illuminations were observed and analyzed by Semantic Differential method. It became clear that factor scores were increased by the illumination from around the display and ceiling.

CT画像中の臓器や組織の形状を3次元的に提示することは診断支援、医療教育などに大変有用であるが、各臓器の表面形状を3次元的に抽出する必要がある。中でも膵臓は周囲を他の臓器に囲まれて境界が不明瞭であるため、抽出が困難とされている。本研究では利用者の知識を簡易に活用するインタラクティブ抽出手法を提案する。初めに利用者の手によりCT画像中から膵臓全体を含む領域を粗く指定する。次に、この領域の境界上の任意の点と、同領域の中心線上の最近傍点とを結び、その線分におけるCT値変化の微係数を計算、微係数の最大値をその方向におけるエッジ信頼度と定義する。このようにして作成したエッジ信頼度分布を自動抽出手法に利用することで不鮮明な輪郭部における形状の漏出エラーを抑制した新しい形状抽出を実現した。本研究ではChan-VeseモデルのLevel set抽出手法に適用し、安定かつ高精度に膵臓形状を得ることができた。(著者抄録)

我々は製造段階での作業に対する質感の変化を視覚的にシミュレーションする質感再現システムを開発している.目標とするシステムには作業と連動する質感再現が重要で,実作業,特に手動作との連携が必要不可欠である.そこで本研究では机上作業における手の位置と動作をロバストに計測する手法を開発した.手の位置の検出には赤外線のみを発するプロジェクタで投影した構造パターンを用い,机上から一定高さ以内の手の位置のみを検出することで,作業時のみの手の位置を検出することができる.また手動作判定に必要な手の形状計測には空間コード化法を用いたが,計測範囲を計算で求めた手の周辺位置のみと限定することで計測及び形状再構築の処理時間を飛躍的に削減することができた.本報告ではこれら画像処理的な手法と,試作したプロトタイプシステムの性能について述べる.

In recent years, various kinds of endoscope have been developed and widely used to endoscopic biopsy, endoscopic operation and endoscopy. The size of the inflammatory part is important to determine a method of medical treatment. However, it is not easy to measure absolute size of inflammatory part such as ulcer, cancer and polyp from the endoscopic image. Therefore, it is required measuring the size of those part in endoscopy. In this paper, we propose a new method to measure the absolute length in a straight line between arbitrary two points based on the photogrammetry using endoscope with magnetic tracking sensor which gives camera position and angle. In this method, the stereocorresponding points between two endoscopic images are determined by the endoscopist without any apparatus of projection and calculation to find the stereo correspondences, then the absolute length can be calculated on the basis of the photogrammetry. The evaluation experiment using a checkerboard showed that the errors of the measurements are less than 2% of the target length when the baseline is sufficiently-long. © 2009 SPIE.

In this paper, spatio-velocity contrast sensitivities were measured by using red-green and blue-yellow grating stimuli. The results show the contrast sensitivities at the spatial frequencies over one cycle-per-degree decrease gradually as the velocities of the stimuli increase.

In this paper, we proposed and validated SV-CIELAB which is a video quality assessment (VQA) method using a spatio-velocity contrast sensitivity function (SV-CSF). The SV-CSF consists of the relationship among contrast sensitivities, spatial frequencies and velocities of stimuli. We used the SV-CSF for filtering original and distorted videos. The criteria in our method are obtained by calculating CIELAB color differences between filtered videos. From the experimental results for the validation, it was shown that SV-CIELAB is the more efficient VQA method than conventional methods such as CIELAB color difference, Spatial-CIELAB and so on.

Rapid proto-typing of product is important technology that the computer science can contribute. Especially, the evaluation of realistic appearance decides the quality of the product in a final process. As the industrial design application, we have been developing the Appearance-based Display System by using the radiance control of projector and technique of mixed-reality. In this paper, we proposed a novel reproduction system of three-dimensional appearance with glossiness, which is controlled by two or more projection images. R. Raskar et al. proposed 2-pass rendering method to consider the 3D geometry. However, view-dependent shading, such as specular highlights is not considered in this method. Because the shading view is assumed to be the same as rendering view or user's viewpoint, the view-dependent shading could not ensure consistency. Therefore, in this paper, we divide 3D geometry into object shape, virtual light direction (shading direction), user's viewpoint (view-dependent rendering view) and projector position(view-independent rendering view). 3D geometry of the viewpoint and the object can be calibrated by the space encoding method with projector-camera system. According to the geometry calibration, the partial responsibility and compensation of the intensities errors such as occlusions, overlaps and attenuation for each projection is automatically decided. As the result, this system can generate the real appearance of gloss, color and shade on the mock object's surface.

In this paper, we have clarified the relationship between ambient illumination and psychological factors in viewing of display images. Psychological factors were obtained by the factor analysis with the results of the semantic differential (SD) method. In the psychological experiments, subjects evaluated the impressions of displayed images with changing ambient illuminating conditions. The illumination conditions were controlled by a fluorescent ceiling light and a color LED illumination which was located behind the display. We experimented under two kinds of conditions. One was the experiment with changing brightness of the ambient illumination. The other was the experiment with changing the colors of the background illumination. In the results of the experiment, two factors "realistic sensation, dynamism" and "comfortable," were extracted under different brightness of the ambient illumination of the display surroundings. It was shown that the "comfortable" was improved by the brightness of display surroundings. On the other hand, when the illumination color of surroundings was changed, three factors "comfortable," "realistic sensation, dynamism" and "activity" were extracted. It was also shown that the value of "comfortable" and "realistic sensation, dynamism" increased when the display surroundings were illuminated by the average color of the image contents. © 2009 SPIE-IS&amp T.

In recent years, the resolution of display devices has been extremely increased. The resolution of video camera (except very expensive one), however, is quite lower than that of display since it is difficult to achieve high spatial resolution with specific frame rate (e.g. 30 frames per second) due to the limited bandwidth. The resolution of image can be increased by interpolation, such as bi-cubic interpolation, but in this method it is known that the edges of image are blurred. To create plausible high-frequency details in the blurred image, super-resolution technique has been studied for a long time. In this paper, we proose a new algorithm for video super-resolution by considering multi-sensor camera system. The multi-sensor camera can capture two types video sequence as follow (a) high-resolution with low frame rate luminance sequence, (b) low-resolution with high frame rate color sequences. The training pairs for super-resolution are obtained from these two sequences. The relationships between the high- and low-resolution frames are trained using pixel-based feature named "texton" and stored in the database with their spatial distribution. The low-resolution sequences are then represented with texton and each texton is substituted by searching the trained database to create high-resolution features in output sequences. The experimental results showed that the proposed method can well reproduce both the detail regions and sharp edges of the scene. It was also shown that the PSNR of the image obtained by proposed method is improved compared to the image by bi-cubic interpolation method. © 2009 SPIE-IS&amp T.