Recently, in computer graphics, not only photo-realistic-rendering but also non-photo-realistic-rendering is researched briskly. In this paper, focusing on pen-and-ink illustration that is usually put in books or used for technical illustration, we propose a method for generating a pen-and-ink illustration from a given gray-scale image. There are two fundamental styles in pen-and-ink illustration: one is line drawing style that represents an illustration by lines, and the other is stipple drawing style that represents one by dots. If the size of a line or a dot depends on the resolution of display device, we have one serious problem; in low-resolution device, the generated illustration can be given an artifact, that is, artificial pattern caused by the connection of lines or dots. By the method proposed in this paper, such artifact can be reduced using smoothness-line, which is given the fluctuation by 1/f noise, and smoothness-dot.

View full abstract

Recently many researchers have been trying to display natural phenomena such as clouds, ocean waves, and flames using computer graphics. This paper proposes a fast method for generating images of clouds. Since clouds consist of particles with a high albedo, multiple scattering cannot be ignored. In our method, the density distribution of clouds is expressed by sets of metaballs. A metaball is a sphere and a density distribution is defined inside it. Multiple scattering is taken into account by computing the energy transfer between metaballs. To reduce the computation time, metaballs are grouped into a hierarchical structure, and graphics hardware is utilized. The proposed method is applied to creating images of outdoor scenes to demonstrate its usefulness.

View full abstract

In this paper, we propose a new method for Image Based Rendering. Many studies using stereo vision techniques, which reconstruct shape of the objects, have been reported in the field of Image Based Rendering. However, these techniques cannot avoid difficulty in determining the positions of all object surfaces explicitly. Based on the thought that the problem of Image Based Rendering is estimation of the color of incident light into virtual viewpoint, our method divides the object space into voxels, and represents the probability of existence of the object as a continuous value calculated from the differences of colors projected onto input images, and directly perform rendering using the technique of Volume Rendering. Consequently, it can synthesize virtual viewpoint images without determining the surface depth explicitly by very simple algorithm.

View full abstract

Line Integral Convolution (LIC) is a promising method for visualizing 2 D dense flow fields. Several extensions of the method have been proposed for visualizing 3 D flow fields. However, volume rendering of 3 D LIC textures has been considered ineffective, because optical integration in depth tends to spoil the coherent structures along local streamlines. In this paper, we propose a selective volume rendering of 3 D LIC textures based on significance maps, which are derived by the cross-sectional vector field topology analysis, and a specific illumination model for 3D streamlines.

View full abstract

We propose a high speed and high accuracy method measuring depth map. In the previous letter, we demonstrated a hierarchical phase shifting method to measure three-dimensional (3D) shape through active vision approach. Here we carry out quantitative evaluation of the accuracy for the proposed method by comparing with the ordinary method. We confirm that the proposed method improves the accuracy and solves the phase unwrap problem by the introduction of hierarchical pattern projection starting from one-cycle sinusoidal grating. In this paper, we consider that the magnitude of spatial frequency, which is incremented with the change of hierarchy lelels, is determined by the error included in the phase estimation from sequential images. Furthermore, we carry out the experiment for several cases of the phase estimation error, and we discuss the relationship of the estimation accuracy and the image frames required for the precise measurement. We also try quantitative experiments to compare the robustness between the proposed method and ordinary one (the space encoding method). As the results, the proposed method is superior to the ordinary method in the smoothness of depth map (about three times) and in the number of image frames required for the measurement (about half).

View full abstract

To realize natural human interface, there are many studies to synthesize facial expressions using computer graphics. In this paper, our goal is to reconstruct a 3D structure of a face with arbitrary expressions (expression-B) using 3D structure with expression-A, two images with expression-A, and domain knowledge of expressions. First, from input data, two projection matrices, which project the 3D structure to two input images, are estimated. Second, deformation from expression-A to expression-B is calculated as the difference vector of corresponding points in the images with expression-A and-B, where the images with expression-A and-B are from the image sample database. By using the difference vectors, similar property of face, symmetrical property of face and geometrical constraint, the images with expression-B are estimated. It two projection matrices and two facial images with expression-B are given, 3D structure with expression-B can be reconstructed. Experimental results show that 3D structure with expression could be reconstructed with fairly good performance.

View full abstract

In this paper, the authors propose a method of identifying individuals using fractal coding facial image. There is a method by using the facial information as one of the methods for identifying individuals. These methods had problems in the sense that precise identifications are difficult unless the direction of faces or the distance between the camera and faces are fixed accurately. On the other hand, the fractal coding technique attracts a great deal of attention of researchers because it is an image compressing technique utilize geometric redundancy in the image. The compressed data with fractal coding is a set of parameters of affine mapping. Therefore, we think that the effective data for identifying the faces is obtained by parameters fracatal coding.

View full abstract

In this paper, we present a new method of a feature extraction and face recognition for 3D human facial models extracted by 3D range scanner. We extract typical feature points from 3D model by combining physical feature parameters such as 3D curvatures and normal vectors with texture features. These feature parameters are stored in facial model database and face identification can be realized by constructing spring surface among these points, calculate spring energy between feature points and database facial models to find correspondance model.

View full abstract

We have developed a face model creation system for expression of persons in 3-D virtual space communication system. The system realizes robust image process in background and illuminance condition for input face image. We compared 3-D facial expression in virtual space with the system with 2-D facial expression in virtual space and face-to-face meeting. We measured that contribution of facial information in an encounter process in communication. As the result of analysis, we found effectiveness of 3-D facial expression in distance and angle at the finding a target person, and time to find the one compared with 2-D facial expression. But it was inferior to face-to-face meeting. Based on the result, we discuss the necessity of the system to be used for the communication system.

View full abstract

Personal handy terminals and computers have come into wide use for digital cartoons through the Internet. There will be a great distribution for digital cartoons in near future and it brings a serious problem of copyright protection because of illegal copy and use. However, we have only insufficient tools for this problem, now. This paper presents a new watermarking scheme of digital color cartoons. We pay attention to the colored region of a picture and introduce a concept of modulo computation for a group of the specified pixels with the same color vector. The watermark bit is expressed with the region of the same color vector. Our experimental results show good pictures and robust watermarks against some attacks.

View full abstract

In this paper, a front end system which uses audio and video information to track the people or other sound sources in the ordinary room has developed. The microphone array is used for determining the spatial location of the sound; the active video camera acquires the image of the area where the sound is detected, detects the people in the image by using skin color and can zoom and track a speaker. Several add-ons to the system include various visualization tools such as on-screen displays of waveforms, correlation plots, spectrum plots, spatial acoustic energy distribution, running time-frequency acoustic energy plots, and the possibility of real-time beamforming with real-time output to the headphones. The system can be used as a front-end for the non-encumbering human-computer interaction by video and audio means.

View full abstract