To display realistic water drops on glass surfaces, we added a new method, that controls the shape change of water drops, to the conventional motion control method. The shape of a water drop was defined as a sphere or a metaball in the conventional method. However, this is too simple and un realistic. In addition, the shape variance based on the time water marks were created by the movement of water drops was not correctly displayed. Thus we defined a water drop as one mainmetaball surrounded by some submetaballs. By controlling the distances of the submetaballs from the mainmetaball, we can display complex shapes. For example, we can give a different shape to each water drop, which can be changed based on its states like generation, movement, union, separation, etc. Since the motion of each water drop is computed from the position and speed of the mainmetaball, conventional motion control methods can be used.
In this paper, we developed a method for sampling and modeling reflectance of 3-D objects and performing real-time rendering of modeled objects. Our method can be used to render modeled objects from an arbitrary viewing direction with illuminations from point lights at any given place or with any illumination map. To achieve this goal, we developed a platform that can sample appearances of target objects with 4DOFs parameters of light and viewing directions. Light-dependent variations of sampled data were approximated using spherical harmonics functions as bases. To render a modeled object, we generated its appearance by synthesizing the effects of illuminations using spherical harmonics functions and by interpolating between sampled viewing directions. The appearance was then texture-mapped to the shape model. Using our method, we could represent reflectance properties of surfaces of 3-D objects that were parametrized by 4DOFs light and viewing directions. We also successfully rendered the modeled objects using programmable vertex/pixel shaders with real-time performance.
3-D video records dynamic 3-D visual events as is. Applications of 3-D video include wide varieties of human activities. For promoting these applications in everyday life, a standardized compression scheme for 3-D video is required. We developed a practical and effective scheme for representing and compressing 3-D video named “skin-off,” in which both the geometric and visual information are efficiently represented by cutting a 3-D mesh and mapping it onto a 2-D array. Skin-off has much in common with geometry images proposed by Gu et al. However, while geometry images use only 3-D surface shape information to generate 2-D images, skin-off uses both 3-D shape and texture information to generate them. This enables us to achieve higher image quality with limited bandwidth. Experimental results demonstrate the effectiveness of skin-off.
In binary phase encoding of computer-generated holograms (CGHs), quantization usually decreases the quality of the reconstructed image. For reducing the quantization and coding noises, we developed an adaptive error-diffusion method by using the simultaneous perturbation algorithm in which the diffusion coefficients are adapted for the size and layout of the object. Furthermore, binary phase-encoded CGHs were fabricated by using a mask-transferring process. Its optical reconstructions and numerical reconstructions in a Fourier setup were demonstrated.
We developed a ray-space coding scheme in which an image-based rendering technique is introduced at the beginning of coding. Our coder first synthesizes an image at an arbitrarily given viewpoint, which is called a representative viewpoint, and then predicts input images by using the synthesized image as a reference. It can produce a scalable bitstream, which enables high-quality rendering around the representative viewpoint even at extremely low bit rates. Our experimental results show that this coding scheme has good coding efficiency for both multi-camera images and integral photography.
We describe a method for aligning multiple range images given by a range finder. We used range images of the insides and outsides of buildings, which often contain many planes. Our method for registering range images is to refine the iterative closest point (ICP) algorithm using not only 3-D information but normal vectors of planes in the scene since structures of buildings often consist of planes. Analysis showed that multiple range images were aligned more accurately and faster using our method compared with the standard ICP algorithm.
We developed a technique for watermarking an MPEG-2 video stream. Our method keeps the bit rate strictly within that of the input stream. Although the need for keeping the rate of a watermarked stream under control has been recognized for many years, most proposals have controlled the watermark embedding at the DCT level. We have analyzed the syntax change due to watermarking and show that the change is closed in the slice containing the DCT coefficient to be modified, although it could propagate beyond the block boundary, sometimes even into successive macroblocks. Based on this analysis, we developed a new embedding algorithm to adjust the bit rate of the resulting stream precisely at the end of every slice. In our experiments, about 30-60% of watermarks were successfully embedded into 6-Mbps (bit/second) video stream without any increase in bit rate compared to unrestricted embedding.
Visual stimuli during motion can induce a sense of physical discomfort. To examine the influence of visual stimuli on the human body, we studied autonomic responses when subjects watched videos of a roller-coaster ride and computer graphics at three different replay speeds. We measured the subjects' heart rates, mean blood pressures, respiration rates, and tidal volumes while they watched the video, and we focused on the relationship between the reaction and replay speeds. As a result, we found that the replay speed influences the heart rate, which changed greatly during a curve scene at double speed. Subjects' respiration rates increased the most when the replay speed was quadrupled during a curve scene.
A field survey and a web questionnaire survey were conducted for obtaining data about television viewing conditions at home. In the field survey, measurements were made concerning the illuminance and luminance of television screens and television viewing distances and angles for 275 viewing locations at 50 homes. In the web questionnaire survey, 290 families were asked about room size, present television screen size, desired television screen size, and viewing conditions. The results showed that the mean screen illuminance was 108 lx, and the mean luminance ratio of black and white was 72. The mean viewing distance was 266 cm, and the viewing-angle range was ±60 degrees from the horizontal, ±30 degrees from the vertical. Correlation analysis of screen sizes revealed that the best predictor of desired television screen size was the present television screen size.
Light reflected from an object to a camera is a mixture of light from specular and diffuse reflection. This has important implications for many computer vision tasks, such as image matching and understanding. Many applications, for example, digital contents production, photorealistic image synthesis, and motion analysis, may require the diffuse and specular reflection to be separated. We present an approach for separating the diffuse and specular components of object surface reflection. This approach is based on the well-known dichromatic reflection model, however it separates reflections from reflectance fields constructed for every point on a 3-D object surface. Our method can prevent having to segment image into several uniformly-colored areas. Our method can thus separate reflection from an object surface with a complicated texture. We analyzed the properties of the reflectance field constructed from original frames and showed how to separate reflection components for each 3-D point. Experiments on real scenes showed that our method was successful.
The conventional robust method for clustering arbitrarily-shaped clusters takes a long time to process. To reduce the processing time, we developed a robust method for clustering arbitrarily-shaped clusters based on labeling by ascending order the distance between clusters.
We fabricated a low-voltage 32×32-pixel pulse-frequency-modulation vision chip with an image processing function executed in a pulse domain in a standard 0.35-μm CMOS technology. We showed the characteristics and feasibility of our vision chip architecture.
Black outs and white outs of captured images usually occur due to shortages of dynamic range. To reduce black outs and white outs, we investigated the applicability of an image sensor with a function for detecting a modulated light signal. The image sensor was fabricated originally to detect light whose intensity was modulated with a certain frequency. The image sensor can also capture two images almost simultaneously. Both bright (long exposure time) and dark (short exposure time) images can be captured almost simultaneously using this sensor. Two images, one of each kind, are appropriately compounded to generate an output image in which the black out and white out is reduced. Experimental results showed that the image sensor can reduce black out and white out. Unwanted effects caused by changes of light intensity during exposure were also drastically reduced.
Colorizing monochrome image sequences is carried out by pasting the color information of the pictures colored manually every several frames, according to motion vectors, which are widely used in frame-to-frame coding assuming “solid object motion”.
During the design process for manufacturing car seats, models of products made from cloth are created by CAD/CAM computer graphics. Among car and parts makers, industrial design using computers is called computer-aided industrial design (CAID), and some examples of CAID have practical use in designing solid exterior sells for cars. However, for models of products made from cloth, few have been put to practical use, because creating images with cloth reflection characteristics is difficult. We developed a method for creating computer graphics of cloth-covered car seats using a bidirectional reflectance distribution function (BRDF) and computer color matching (CCM).