It is difficult for amateurs to set up a proper camera path in a 3-dimensional computer graphics (CG) space. The viewing condition also changes the proper camerawork. For example, a large viewing angle and screen size provides a dynamic impression. We studied the relation between the camerawork and viewing condition for this paper. As a result, we found that slightly curved paths enriched the quality of expression, and linear or very curved paths provide interesting and dizzy impressions at large viewing angles and big screen sizes. However, acrobat camerawork should be used when the screen size is small. These discussions state that controlling the camerawork is useful for creating video contents adaptable to specific viewing conditions.

View full abstract

Traffic sign recognition systems can be used to assist drivers and improve road safety. The system is expected to recognize traffic signs at even greater distances in order to give drivers as much warning as possible based on the road conditions. A hybrid camera system is proposed in this paper with the goal of increasing the recognition distance compared to the conventional systems. In this system, an active telephoto camera is used as an assistant to a wide angle camera. Traffic sign detection and classification are processed separately for the different images from the wide angle camera and telephoto camera, respectively. The image from the telephoto camera provides enough information for a classification when the resolution of the detected traffic sign is low from the wide angle camera. The experimental results demonstrated that the recognition distance of the proposed system is improved compared to the conventional systems.

View full abstract

We present an augmented reality (AR) system that is based on a tabletop system and has a hemisphere omnidirectional camera. We perpendicularly set the camera on a tabletop display and showed an omnidirectional image on the display screen. When users present objects to the camera, these objects are captured and recognized by using a specific object recognition technique. On the bases of the recognition result, AR contents are overlaid onto the omnidirectional image on the screen. Our system includes following two advantages. First, users can interact with each other by recognizing the relationship among the users, objects, and their AR contents because the entire surrounding of the tabletop is shown as a circular image on the screen. Second, multiple users interact here without the need for any specific devices. To describe the applicability of our system, we present various examples of applications exploiting its advantages.

View full abstract

Recently, the government of Japan has been working on creating a society in which people enjoy participating in sports whenever, wherever, and at any age. To create such a society, there needs to be more players of all kinds of games. However, at present, more than half the public and private gyms in Japan lack instructors. In this research, we focused on the game of tennis doubles and have developed a system for improving tactics using a digital video camera even when no instructors are available. The position of each player when stroking the ball is obtained to track the player and the ball in a dynamic image. A tactic that is right for the situation and based on the positional relations of each player is presented to both the offensive and defensive side. This system can be used to encourage more people to learn to play tennis even when they cannot learn from an instructor and to build a lifelong sports society.

View full abstract

We have proposed a method for clipping a rectangular region from an image by minimizing a weighted intersection of two color histograms that are constructed with pixels inside and outside the rectangular region. We propose a method for detecting salient regions which is based on two color histograms obtained by the clipping. We experimentally compared our method with the latest method by Achanta et al., and the results show that salient regions can be detected more clearly with our method.

View full abstract

In non-photorealistic rendering of stained-glass images by Voronoi tessellation, the preservation of input images is improved by extension of isotropic Voronoi tessellation to anisotropic Voronoi tessellation. In this paper, we show that the preservation of input images in this anisotropic Voronoi method is further improved by extension of the point arrangement from isotropic Voronoi distribution to an anisotropic one.

View full abstract

We describe a method for reducing an image by seam carving using a color saliency map in which higher values are assigned to the pixels with the object's colors. Our experimental results show that the proposed method can preserve the salient regions in the reduced images better than the conventional methods.

View full abstract

A fog display can project images in free space. However, as the screen is planar, it can represent only plane images. We thus propose a fog display that represents stereoscopic images by providing a physically 3D-shaped fog screen. We implemented a dynamic duct control mechanism, and obtained a three dimensional shape of the fog screen, which represent depth of target objects. We show our results and the details of our proposal here.

View full abstract

Image projection technology has been studied that gives an immersive feeling by projecting images around a viewer. Immersive projection displays are large-scale and expensive systems. This paper proposes a technique for small-scale and inexpensive immersive projection. We used a fish-eye lens to expand the projected plane and adjusted the expanded plane to the walls based on our geometric correction.

View full abstract

An algorithm that improves the robustness of an ear biometric system is proposed and targeted for use in surveillance systems based on ear biometrics. To deal with pose variation when ears are rotated in depth, Gabor jets of different poses are estimated and used as training data for a discriminant-analysis-based classifier. Experimental evaluations show the effectiveness of the proposed algorithm and the possibility of improving the robustness of a single-view-based ear surveillance-system.

View full abstract

We investigated the educational effectiveness of a ‘distance lecture’, i.e., one based on instruction written on a blackboard and displayed to students through IP networks. We have identified the factors that affect the effectiveness of the system in a multidimensional way. Educational effectiveness is evaluated here in terms of satisfaction and understanding. We investigated the satisfaction in a multidimensional way by the semantic differential (SD) method with many adjective pairs. The understanding is expressed by the learning quotient, which is defined as the ratio of correct answers in the pre-test to those in the post-test. We also evaluated the level of the teacher's satisfaction. We found that the level of satisfaction of the students closely correlates with the quality of audio and video, which affects how the teacher's presence is perceived by the students. We also noticed that the learning quotient has a strong relationship to the content of the lecture and to subjective understanding. In addition, we see that the teacher requires a higher response of the system than that required by the students.

View full abstract

Facial information is a principal component in nonverbal communication, and it gives a wide variety of impressions to the observer. If we can clarify the relationships between facial information and impression, we would be able to easily input the right visual information on a face and thus manipulate impression given from a facial expression. We focused on the shading on the face. Although the ability to control the perceived shape by manipulating shading, e.g., through makeup, qualitatively evaluating this has not been easy. To find a way to do this, we developed a system for editing shading on a face by using shading generated from a target model with a known 3-D shape. We have also developed a qualitative method to evaluate the relationship between the shading and the perceived shape. We first describe the system in detail, and then demonstrate how it is effective at providing both qualitative and quantitative evaluations from real images.

View full abstract

We have developed a projector which can display high-resolution images of Super Hi-Vision (7680 x 4320pixels) and has a high dynamic range of 1.1 million to 1. This projector features a serial combination of two modulation blocks: the first for chrominance and the second for luminance. While a conventional projector has one modulation block that contains three devices to modulate red, green, and blue light, our projector has another modulation block that contains a device for luminance to further modulate the combined RGB modulated light. We measured some characteristics of the projector, which are modulation transfer function, dynamic range, and tone reproduction. We found that the projector can resolve Super Hi-Vision images and that it has an extremely high dynamic range and a fine 12-bit tone reproduction. We did two types of subjective evaluations with the projector to research the effect of the resolution of the chrominance component and that of the limited dynamic range of the display on the total image quality. The results of the first evaluation indicate viewers do not notice the image degradation when the resolution of the chrominance component is degraded to a half of that of the luminance component. The results of the second evaluation indicate the dynamic range of at least 1000:1 is necessary for displaying natural images and the dynamic range of about 30000:1 is necessary for displaying special images which have low average picture levels.

View full abstract