Stereoscopic view with binocular disparity produces a distorted spatial layout, such as, the puppet theater and cardboard effects. In this paper, we propose a disparity mapping method for reducing the distortion of 3D images. The method is divided into two parts. First, we defined the perceptual model and the modified spatial layout based on this model. Second, the differences in adjacent disparities of the objects were interpolated. We conducted a subjective test to evaluate the proposed method when applied to 3D images taken with a smartphone equipped with a twin-lens camera. There was a significant difference (p < 0.01) between the original 3D images and those of the proposed method. The results of our experiments show that the proposed method makes 3D images more round and gives them a sense of depth.
We are conducting research on a wireless link system that is capable of achieving high link reliability over a wide area for use in mobile relay broadcast programs such as road races. For such systems to be feasible, we developed a macrodiversity reception system using many distributed remote antennas and radio-over-fiber links. Our receiver combines four diversity branches with good signal quality based on the minimum mean square error (MMSE) algorithm. We describe the MMSE weight vector calculation using fixed-point operation for hardware implementation, prototyping of a macrodiversity receiver, and its use in an actual mobile relay broadcast program. Our system was used as a terrestrial reception system of high-definition videos shot from a motorcycle in the Kyoto Ekiden relay race, and achieved highly stable transmission without outage for half the course.
We have many chances to watch stereoscopic images, i.e., 3D movies and 3D TV programs. However, there has been little focus on the relationship between screen size and viewing distance with regards to stereoscopic images. We explain a subjective evaluation of the preferable viewing distance (PVD) and minimum viewing distance (MVD) for both still and moving images of different sizes, including monoscopic and stereoscopic images with different displarity maps. The results indicate that (1) for both the PVD and the MVD, the people stood closer to the images relatively as the image sizes became larger in terms of absloute display height, (2) for smaller image sizes, people stood farer to the stereoscopic images than monoscopic images, either the PVD and the MVD, and (3) differences in the disparity map in stereoscopic images do not affect the viewing distance in either still or moving images. These results may provide a guide to visual environmental design of larger stereoscopic display systems.
We propose a method for generating halftone images by scaling Voronoi regions in centroidal Voronoi tessellations (CVTs). We first segment a given image plane into relatively small Voronoi regions by using a CVT, and reduce the area of each Voronoi region in proportion to its mean brightness. That is, the brighter a Voronoi region, the smaller it becomes. Finally, we draw the reduced Voronoi regions in black on a white image plane to obtain a halftone image. Experimental results show that the generated halftone images faithfully reproduce the continuous tone in the original images and retain the structure of CVTs simultaneously.
The embossing of bronze plates is a type of engraving art, created by hitting a nail onto a bronze plate with a hammer. The final appearance critically depends on several factors such as the nail gauge and hit-point density. It is necessary to adequately determine these factors before starting the process because the process is irreversible. This is not generally easy for novice creators. We propose a computer simulation system for embossed bronze plate creation, which allows users to estimate appearances by changing factors such as nail gauge. In a user-test, subjects reported that their creation process became much easier with the proposed system.
We propose a non-photorealistic rendering method for generating sand picture-like images from photographs. The proposed method is performed in two steps. First, images with flow patterns created from finger tracing are generated with an anisotropic filter. Second, sand-pattern images are generated with a new filter called "sand filter" which incorporates the idea of iris filters. The new filter has a characteristic emphasizing images including low contrast parts. Therefore, it is possible to generate images like sand picture drawings.