We propose an intuitive point cloud modeler which can operate a three dimensional model by hand gesture. The system recognizes the gesture by measuring hands using Kinect and transforms the point cloud data in accordance with the hand gesture. In addition, we achieve the high speed processing by using GPU.

抄録全体を表示

Aiming an automatic method for creation of pixel art, four Imaging processes, Lanczos-Resampling, MDA-Clustering, Canny Edge Detector, and Coloring of Contour process were used. The last process is particularly new proposal of pixel art and results in an expression almost like 16-bit video-game console.

抄録全体を表示

We propose a computational model to generate life-like motion for firefly-like creatures (Virtual Fireflies). By using a two-stage stochastic model and simple operational elements, we could generate various life-like motion patterns.

抄録全体を表示

To find and treat diseased part of intracerebral, doctors attempt to understand sterically intracerebral activity from cross-sectional view. This paper proposes a projection mapping method for representation of the intracerebral activity, that is achieved by making the 3DCG based on the image of PET and MRI.

抄録全体を表示

In this paper, we propose a free-viewpoint video generation method for sports scene using CG models as well as natural textures. Our proposed method detects occlusions and replaces the texture of each occluded object by the corresponding CG model seamlessly. Experimental results show that our proposed method improves the image quality of generated free-viewpoint video.

抄録全体を表示

We have studied a modeling method to generate 3D character object. Simplified method was proposed and an experimental result showed its effectiveness.

抄録全体を表示

This paper proposes an automatic generation method of marker designs using genetic algorithm. The method aims to create a marker which has high-visibility and large amount of information. An accuracy of decoding is also considered in making the valuation.

抄録全体を表示

Recently, a technique of visual cryptography which reveals a secret image by overlaying two other images has attracted attention. As an application of this technique, we propose a method to embed given QR codes in halftone images which can be recognized as the same natural image.

抄録全体を表示

This paper proposes "UtteractiveWall" which attracts people and stirs up them to join what happens there. UtteractiveWall catches the eyes of passersby with bubbles floating on its large-and-wide display, when the person approaches the wall and touches one of the bubbles, a recorded someone's voice is replayed.

抄録全体を表示

Recently research of recognizing real-time motion of a finger in the air to perform an input operation has been carried out. However, because of hand trembling, disturbance and sensor resolution, stable operation is difficult. Therefore, it is necessary to realize stable operation by smoothing the trajectory of the finger in real time. In this study, giving a priority to the smoothness of the trajectory and the delay amount, we compared several smoothing filters to find the best filter to solve the problem.

抄録全体を表示

This paper presents the evaluation results of education contents with 3-D Character. Experimental results show that 3-D Character with highlighted motions or intense emotions is effective to improve student's comprehension.

抄録全体を表示

A dynamically configurable presentation system is presented. In contrast to existing one-way and slide-byslide style such as PowerPoint, the progress of this system is continuous and changeable on-the-fly according to the audience's reaction and understanding. A web-browser-based prototype is extended to handle multiple branches of the progression without showing the branches to the audiences.

抄録全体を表示

We developed Tablet PC based e-learning system, which can learn and study the interesting contents with digital notebook. This system also enables to communicate between teacher and students with web cameras. However, it is difficult to understand the centered state of student for e-learning through a web camera. Therefore, we propose an estimation method for state of observers by using the pattern recognition of their facial expression. The result of this research indicates the relational analysis of the centered state between facial expression and brain wave.

抄録全体を表示

We have developed an automated musical performance system which supports mutual performance exchange between multiple real users and agents in a computer. In this paper, we report the method for following the user's pitch and volume of the voice in real time, based on the information of score.

抄録全体を表示

I have already built a prototype of an automated music composing system named "MUSCAT (MUSic Composer bAsed on color informaTion)." By choosing one color, my system can provide several appropriate musical components, and compose pieces of music by combining them. In this paper, I describe the outline of a method of musical composition of my system.

抄録全体を表示

Survey errors if they appear on experimental results by any factors by comparing the experimental results of three.

抄録全体を表示

The proposed method inputs characters by touch gestures. Each gesture is defined by motions sliding index finger along edge of the display of a micro size touch device. Start position and length of the gestures are optimized to the romaji input. The method fewer occupies screen than the conventional methods.

抄録全体を表示

Our method is designed to input characters without seeing input device by one hand. Each character is specified by a touch gesture, which is a combination of vertical and horizontal movement of a thumb. Position of the thumb is ignored. Input guide that assists choosing gestures is also installed.

抄録全体を表示

Multimodal synchronization is essential for virtual agents. This paper presents an effective approach to synchronize gesture with speech by motion graph technique. The basic idea is to consider the synchronization problem as a motion synthesis problem, where the goal is to match the duration and timing between the gesture and speech.

抄録全体を表示

We verified usability improvements by introducing sound localization for voice picking. With the introduction of Head Tracking (HT), task completion time was reduced about 1.8 seconds compared to diotic conditions. The results of subjective evaluation, showed that the HT condition is preferred over diotic condition. It can be said that with the introduction of sound localization in voice announcements, user satisfaction and work efficiency is improved.

抄録全体を表示

Social TV services, which consist of Video on Demand and text based user's communication, should be designed by taking into account the users' behavior. We conducted a field trial of the social TV system called "teleda" and analyzed the user logs. The results show that active communication users who like to communicate amongst each other prefer services reflecting other users' behavior, and passive communication users prefer services based on their preferences and context.

抄録全体を表示

A method for predicting viewers' ratings of video content on the basis of viewers' behaviors in a typical home environment is proposed. It detects whether the viewer is gazing at the video content or not on the basis of the viewer's head pose. The simulation results suggest that the proposed method can robustly predict viewers' ratings of video content.

抄録全体を表示

We have developed an auto-detecting TV program system based on three algorithms; detecting program names, comparing text of the tweets with program's meta-data, and comparing time stamp of the tweet with the TV timetable. The system performs an overall assessment based on the candidates put out from those three algorithms, and is able to determine not only the name of the program, but also individual broadcasting.

抄録全体を表示

This paper describes a new color-name analysis method on images and its application one for some purposes. This method is based on human visual perception. This helps color vision defectives to discriminate various colors.

抄録全体を表示

Burst video noise like salt-and-pepper is occurred in HDTV signal. For this noise detection, the new method applying to pattern recognition is proposed. It is estimated one threshold value calculating from average and variance to near three pixels and evaluated to be over this value to ups and downs pixel.

抄録全体を表示

This paper describes the technique about 15.6m wide x 3.4m tall, 9,984 x 2,160 pixels, single panoramic sequence projection by synchronized triple 4K projectors lined up horizontally. Images captured with combination of synchronized quad 4K cameras lined up horizontally produce the final single images on screen through several process as panoramic composition, color grading, blending projection.

抄録全体を表示

In case of color hologram, GS(Gerchberg-Saxton) algorithm takes much calculation time. Therefore, we have devised a fast method of speckle noise reduction of color hologram using the color space conversion from RGB color space to YCbCr color space in which the down sampling of the color difference component is used.

抄録全体を表示

In this paper, we propose a video synthesis method for autostereoscopic display. The proposed method acquires minimal and maximal disparity values during a shooting stage and adjusts disparity values by trimming videos. Experiments revealed that the proposed method can provide more comfortable stereoscopic viewing than a conventional method.

抄録全体を表示

A real-time and wide-field-of-view image pickup system for coarse integral volumetric imaging (CIVI) is realized. This system is to apply CIVI display for live action videos generated by the real-time 3D reconstruction. By using multiple RGB-D cameras from different directions, a complete surface of the objects and a wide field of views can be shown in our CIVI displays.

抄録全体を表示

In this paper, we introduce a new type of time-division multiplexing parallax barrier that is laid like "red-green-blue-black". This kind of barrier pattern turns out to be able to present balanced auto-stereoscopy where crosstalk and stripe noise are weakened to a low level at the same time.

抄録全体を表示

Display devices capable of presenting stereoscopic images have become very popular, but the color gamut of the device is insufficient because of using the three RGB primary colors. This study propose a three-dimensional image display device wide color gamut.

抄録全体を表示

This paper proposes a method of 3D shape measurement for translucent object. A translucent object is difficult to measure the 3D shape because of its permeability which causes light scattering. Therefore, this research is aimed at measuring 3D shape of translucent object by reconsidering existing method.

抄録全体を表示

We have clarified problems of making 3D synthesis images. An effective method has been described by using two virtual cameras operated by TVML(TVprogram Making Language).

抄録全体を表示

We propose an automatic segmentation method and a human depth correction method using contour definition. It can make precisely human depth and natural 3D images within the human who put up his hands, wear big hat and so on.

抄録全体を表示

In this paper, we produced integral 3D contents with camerawork of TV programs using multi-viewpoint robotic camera system. To produce the integral 3D contents, we estimated camera parameters of robotic cameras using all frames of multi-viewpoint video. 3D models of subject were generated using the camera parameters and converted to integral 3D images.

抄録全体を表示

We are investigating integral image systems with high resolution. The pitches of lenses and pixels are important parameters for the integral system. In this paper, we proposed the method of converting integral images to stereoscopic images. The subjective evaluation experiments show the proposed method is effective to exam the parameters for integral image.

抄録全体を表示

This paper presents a subcategory method for pedestrian Recognition. Boosting is performed to test data using study data. Clustering divides study data into a subcategory. Improvement in a pedestrian recognition rate is expected by using the study data divided intothe subcategory.

抄録全体を表示

Crack detection is particularly important for safety and cost-effective maintenance of road structures. However there are many difficulties in image-based crack detection for various reasons. In this paper, we propose the novel hybrid method using the homogeneous and percolation model.

抄録全体を表示

Recently, the number of traffic accidents and injured people has been decreasing. However, the traffic-related pedestrian death rate in mode of transportation is increasing. Therefore, pedestrian recognition as safe driving support system has been studying. In this study, we discuss two dimensional HOG feature based polar coordinates as image feature in pedestrian detection.

抄録全体を表示

This work focuses on a human identification by measurement of the joint of the fingers and principal component analysis and proposes a method. In the proposed method focuses on a individuality of the joint of the fingers and interphalangeal joint and measures them. Experimental results show that the proposed method can identify efficiently.

抄録全体を表示

There are many transparent or translucent objects in real world, for example glasses and plastics. However most of image recognition techniques use the object color information and these can't apply to transparent objects. This paper proposes a method for detection of transparent objects from image using texture analysis technics.

抄録全体を表示

This paper proposes a method of recognition of object colors under varying illumination condition in videos. The recognition function is the categorical color perception model using a neural network. The object regions are detected using the depth information from RGB-D camera.

抄録全体を表示

This paper describes an algorithm for automated motion analysis of axonal transport. In this paper, we propose a method which consists of local tracking using kymograph and particle location images, and global tracking based on genetic algorithm , in order to obtain particle tracking with high sensitivity and accuracy.

抄録全体を表示

We propose a fast vision-based pose tracking method, which is suitable for mobile augmented reality. Conventional pyramid-based approaches have been faced with the problem of complexity since their processing pyramid level is fixed regardless of camera motion state. Our method estimates the motion state initially from the movement of a subset of keypoints to be tracked, and switches processing pyramid levels according to the motion state. Experimental results demonstrate that our method speeds up the conventional pose tracking by about 37% without loss of robustness and precision.

抄録全体を表示