Ocean Acoustic Tomography is attracting high attention because of its usefulness for oceanographic survey. However, as the available information is insufficient, errors from the calculation are unavoidable with the past algorithm. Therefore, the new method utilizing multiple wavelets model called WPR (Wavelet Part Reconstruction) method is proposed. In this method, wavelet sampling model, multiple wavelets sampling, and SVD (Singular Value Decomposition) are used. The WPR method can be applied to the image corresponding to a large image size with high resolution using smaller volume of projection data. Then, we evaluate the reconstruction error and the calculation time by the change in the sampling function, the image size, and the number of sensors.

View full abstract

In order to provide a realistic CG model, it is important to include specular reflectance in addition to textures. This paper describes a method for estimating specular parameters for a free-form object with known geometry data using readily available equipments: a PC, a video camera, and a lighting device. First, the object surface is classified into several regions that have identical specular reflectance. Secondly, specular parameters are measured by moving the camera. The positions of the camera and lighting device are estimated based upon pictures from the camera. The system was implemented on a PC. The total time to measure optical properties at 26 frames (or camera positions) was found to be approximately 30 minutes.

View full abstract

This paper presents a two-step algorithm for detecting video shot boundaries in a wavelet transform domain. It captures gradual shot transitions such as dissolves and fades as well as abrupt shot transitions such as cuts. At the first step, it computes a distance between video intervals to find an isolated interval. Then, at the second step, it computes a distance between frames to specify the exact location of a shot boundary. Its effectiveness is evaluated in terms of recall and precision.

View full abstract

In developing an effective method to detect arbitrary two-dimensional shapes, it is an important research subject how to precisely extract tangent information at each edge pixel from a boundary image. There have been numerous tangent-extraction methods until now, but most of them cannot extract tangents well because they are based on the differential operation of a local 3×3 mask and then they are strongly influenced by noises or sampling errors of an input gray image. This paper proposes a new method to extract tangent information directly from a boundary image. In our method, we fit an analog line or a circle into local edge points by using a Hough-based voting approach and get parameters of its equation. And then, we update the parameters by using an M-estimation technique known as a robust estimator and determine a tangent value. Experimental results show that our method performs effectively and stably even if there exist random noises or discontinuous parts in boundary images. Furthermore it is applicable to edge parts having comparatively large curvatures.

View full abstract

In this paper, we define a face plane to a three-dimensional data set of a facial surface and propose a method to capture as two-dimensional data effectively a change of facial expression by using the face plane. We think that such data obtained from a three-dimensional face data set by mapping the three-dimensional face data to a plane located uniquely and stably is more useful for facial expression analysis than the raw three-dimensional face data. The face plane defined in this paper is solved uniquely and stably from a three-dimensional face data set and is located at the position in a head where normal lines to a face surface are focused. First, we give two definitions of the face plane and present a derivation method for each definition. Second, we propose a facial expression differential map which visualizes a facial expression change from a normal face to a smiling or wry face as analyzing method for facial expression. Finally, by a recognition experiment of facial expression, we demonstrate the effectiveness of the face plane and the facial expression differential map for facial expression analysis.

View full abstract

In this paper, we propose a system for displaying digital contents onto a white real object with an arbitrary shaped surface. This system consists of the uncalibrated camera and projector. We regard the object surface as many small rectangular regions. By calculating homographies between each image planar suface and a devided rectangular region, the projector renders the image which is projected onto the object surface without distortion. The process is performed by using only 2D images in this system, so a user does not need any equipment or special knowledge for 3D measurement of the object surface. In the experiment, we demonstrate that the digital document can be displayed on the surface of a notebook as if it was printed on a real paper surface of the book.

View full abstract