This paper describes two different approaches to the problem of face detection and recognition. One approach determines locations of right and left eyes, nose, and mouth in an input image using Maximum Likelihood detection, and, then, recognizes the obtained canonical face image using a simple distance classifier based on the eigenspace method. The other approach applies the global affine transformation (GAT) correlation technique to matching of the input image and enrolled face templates. The GAT correlation technique simultaneously detects and recognizes a face candidate region by determining the affine-invariant correlation value between two images. We show successful experimental results made on 300 faces × 8 images extracted from the public HOIP face image database.
Individual identification technique, if successfully developed and implemented in the field of medicine, will certainly reduce various sorts of burdens on practitioners, social workers and especially patients who cannot speak,
communicate on move. We fabricated a fully automatic fast face recognition optical parallel correlator (FARCO) (1000-4000 faces/s), based on VanderLugt Correlator. Low error rate of 1.3 to FMR and 2.6 to FNMR were achieved. For the care of moving images, face recognition algorithm for FARCO using a temporal image sequence is proposed. In this paper, we apply this proposed algorithm to dramatic changes in face of children and babies in hospital, using the simulation. As a results, the system proved its capability for identifying moving images.
Originally sound is an object for hearing. However, it becomes possible to elucidate the situation of sound or a sound environment's situation better by catching sound with other sense organs than ears. In this paper, The effect of absorbing sound of a variety of resonance type noise controllers is measured. And the resonance equipment made by transparent acrylics is installed between the laser head of a scanned type laser Doppler vibrometer (Scanning Vibrometer, Following SVM) and a hard wall, we observe multi-points of the hard wall used as a virtual projection and take the data, at last we picture-ize a part for a round term of a resonance signal for every phase to display the sound field.
It is possible to control color light emitted from a projector by using a personal computer, which can specify the chromaticity coordinates of each pixel. Such a system is used as a flexible light source in fields including visual information processing, and computer vision. The DLP (Digital Light Processing) projector consists of a high-speed rotating wheel with four color filters and a set of digital micro-mirrors. Therefore, color control for the DLP projector is not easy compared with the other display devices such as CRT and LCD. This paper proposes a method for controlling color of this projector based on a neural network technique. Moreover we present a method for correcting the emitted radiance on the spatial location of a screen.
We have been studying on measuring, recording, and rendering a painting for the purpose of digital archive for art paintings. So far, a method was developed for estimating the surface properties for oil paintings with gross and specularity, and for determining a 3D light reflection model so that realistic images of the paintings were rendered. The present paper deals with water painting as the art paintings, the surface properties of which are essentially different from the oil paintings. The surface has no gross, but has roughness based on the drawing paper itself. These characteristics are taken into account in digital archiving of water paintings. We use a multi-band imaging system for measuring an paining, and estimating surface normal and spectral reflectance of the surface. The Oren-Nayar model is used as a reflection model. Finally images of the painting are rendered using the estimated surface properties.
The phenomenon of mutual reflection is often seen in our every day lives. When two objects are placed closely, the influence of mutual reflection can be observed between the two objects. This phenomenon causes a problem in the fields of computer vision and image processing because it changes appearance of the scene. This paper describes the spectral analysis of mutual reflection between two surfaces with perfect diffuse reflection. We model the physical phenomenon of multiple reflections of light between two surfaces. We use the spectral imaging system consisting of a multi-band camera. First, we observe the spectral reflectances with the influence of the mutual reflection, and estimate the original spectral reflectances inherent to the object surfaces. Next, we extend the radiosity algorithm to the spectral version and render images of diffuse mutual reflection based on the estimates of the inherent spectral reflectances.
In the field of bioscience, molecular graphics programs have been used for viewing of the three-dimensional (3-D) structure of DNAs and proteins. A detailed technical report about molecular graphics programming was rarely published. Therefore, as a trial of molecular graphics programming, programs based on the graphics libraries such as OpenGL, Mesa, Java3D, Cortona and DirectX were developed using various computer languages corresponding to these libraries. The developed graphics programs have the features of automatic reading of PDB (Protein Data Bank) files and 3-D display of PDB data with the so-called Spacefill and Cartoons representation, and run on MS-Windows and X-Window systems.
To accurately render translucent materials such as marble, milk, and human skins, it is necessary to take into account the subsurface scattering of light. However, rendering translucent objects with the subsurface scattering takes long time for computing. Several methods have been developed to render translucent objects that are represented by polygons. In recent years, point-sampled geometry has been paid to attention as a new rendering primitive that can efficiently represent the object of complex shape. However, there are few methods for rendering translucent objects represented by points. We present a fast calculation method of the subsurface scattering for the point-sampled geometry. In our method, points are stored in an octree data structure. The intensities of the subsurface scattering of light are calculated at eight vertices of each octree node. The intensities of points that represent the translucent object are interpolated by the intensities of vertices of the octree. Our method can render about 300,000 points with the subsurface scarttering at interactive frame rates.
PDAs (Personal Digital Assistants) are usually used on a hand in unstable condition. In the case that beginners at drawing using PDAs want to make illustrations, it is difficult to draw beautiful curves because of unintentional movement of the hands. In this paper, we propose a support system for beginners to make beautiful illustrations using a PDA. By the system, users can easily draw beautiful curves with no unintentional fluctuations. Drawn curves can be corrected locally, and the correction is done by selecting the best curve segment among candidate curve segments inputted by users. From the result of experiments using a prototype of our system, we confirmed that our system was effective for beginners.
Recent surge in interest for 3D shape similarity search and retrieval of 3D shape models has produced many 3D shape similarity comparison methods. Yet, the performance of each shape comparison methods is not sufficient, and the performance varies depending on the kind of shape sought for. In this paper, we propose a systematic approach that combines multiple distances obtained from multiple shape features to improve overall 3D model retrieval system performance. We linearly combine normalized distances by using (1) static weights, and (2) dynamic weights that are determined using a pre-classified training database. As the shape features to be combined, we used (1) four basic shape features (D2, AAD, SPRH, and LFD), as well as (2) their respective multiresolution versions produced by using Ohbuchi and Takei's method. Experimental evaluations showed that the combination of multiresolution SPRH shape features outperforms the LFD, arguably the best performing shape feature today. Furthermore, some combinations of multiple shape features outperformed the LFD by a large margin.
Proliferation of 3D shape models has prompted study into shape similarity search retrieval of 3D models. Descriptors for a 3D shape model need to be compact in order to reduce its storage cost and computational cost for shape similarity comparison. On the other hand, a shape descriptor having better retrieval performance tends to have a large dimension. In this paper, we explore an approach to reduce storage and similarity comparison costs while maintaining retrieval performance. Specifically, we evaluated the effects of principal component analysis (PCA) and its non-linear variant, kernel-PCA, for their effectiveness in reducing dimensions of shape descriptors without sacrificing retrieval performance.
Owing to the spread of digital content, the interest in the information hiding technology such as steganography and digital watermarking has been growing in recent years. Though, very few attempts have been made at the information hiding method for color images in consideration of human color perception. Then we propose a new information hiding method for color images in consideration of human color perception. The number of embedding bits is changing due to the value of RGB component in our method. It is confirmed by subjective test that using our method we can embed more secret information with minimum deterioration than the methods in that secret information is embedded for each pixel evenly.
The digital welfare map is performed as practical activity in Atsugi IT consortium. The digital welfare map can provide welfare information such as the obstacle and life information on the Internet, by which the senior citizen and physically handicapped persons can go out safe. This report shows the outline and its research organization of a system of the digital welfare map first. Next, the safety route search function in the digital welfare map was considered, and it proposes the technique of the calculation of the coefficient of the obstacle uses man's subjectivity. The proof experiments of these were conducted in January, 2005 at the end, and effectiveness was confirmed by the evaluation of the actual people in wheelchairs, and the problem in the digital welfare map in the future was picked up.
We have developed Analog to Digital converting system using image processing for telemetring of oil thermometers and pressure meters. Practical use of this system may realize very effective integration of analog meters to digital based control system.Now this system is evaluated as a viewpoint of application to Automatic surveillance systems in which as yet many traditional analog meters are used.
Now, sign language is used as a conversation language about the deaf person in daily life. Recently, sign language is spreading gradually through the drama of which theme is sign language, the sign language explanation broadcasting, and the sign language news. Along with becoming strong a barrier-free intention in recent years, the concern for sign language that can achieve making of communications barrier-free has improved. However, a healthy person who can use sign language is still little. And sign language has not been approved as the communications means between the healthy person and the deaf person. For achieve communications that using sign language, I clarify the presentation moderate speed of motion and the reading speed according to the length of sign language.