A CMOS image sensor targeted for scene understanding imaging, whose focal plane is split into multiple macro-pixels consisting of 128(H) × 128(V) pixels is discussed. The proposed architecture provides fine and aggressive adjustments of the imaging parameters on the focal plane by way of the independent control on each macro-pixel. As a result, extended intra-scene dynamic range imaging is achieved without reducing the frame rate, which is difficult to achieve with the conventional batch-scan-imaging scheme. Also, the adaptive time resolution control enabling a faster frame rate for moving objects and a slower rate for the background makes it possible to drastically reduce the data transmission bandwidth without compromising the object tracking performance. Based on this concept, we fabricated a test chip using a 0.18-μm 1P5M standard CMOS process. The independent exposure time and frame rate controls were demonstrated with <0.04% column fixed pattern noise (FPN) and </0.11% temporal noise.
Physical models and algorithms for device simulation for high-speed image sensors are proposed. To obtain consistent basic equations for both the device and wave optical simulations, the electromagnetic scalar field is newly introduced to realize current injection and absorption at electrode surfaces and carrier generation-recombination in semiconductors. The propagation of an electromagnetic field induced by electrodes can be considered by using the proposed equations and similar algorithms as used in the FDTD method in device simulation. The 1D field propagations in silicon and Si3N4 are calculated by ignoring magnetic field. It is shown that the proposed equations and algorithms can realize device simulation for high-speed image sensors by considering field propagation in semiconductors.
Driver's visibility diminishes severely in rainy conditions at night. The glare on wet road surfaces causes this poor visibility. In this paper, we propose an image processing method that detects the glare on wet road surfaces by Using multiple onboard cameras. We also propose an image generating method that reduces the effect of glare. By Using our method, glare-reduced images can be obtained and by displaying these images on the monitor equipped in the vehicle, the number of traffic accidents can be reduced. The proposed method is based on the characteristic that the road plane has a common planar surface in multiple camera images. By Using this characteristic and a homography matrix, the glare can be detected. Experimental results obtained by applying the proposed method to real image data show its high performance.
A wide dynamic range image recording technique using a digital micromirror device (DMD) camera with a spatio-temporal exposure pattern (STEP) method is proposed. By incorporating DMD technology, each pixel in the DMD camera can be recorded at a different exposure time. The wide dynamic range images are reconstructed using the spatial and temporal intensity information simultaneously. Preliminary experimental results showed that the DMD camera can record wider dynamic range images with a minimum reduction in resolution.
We have studied holography for photographing and displaying 3D images of real scenes. Although digital holography is known as hologram generation of real scenes, photographing the interference fringe of light has to be performed in the dark room. We propose hologram generation by using a color image and depth map of the same viewing point. In the case of object beam calculation using the maximum diffraction angle depending on wavelength and pixel pitch, we could only observe 3D images in the range of horizontal to 5.6 degrees. In this paper, we report a method that can increase the viewing-zone-angle by using diffraction maps propagating in three directions. As a result, by evaluating the generated hologram with an optical system by the time division, we could easily observe 3D images in the range of horizontal to 16 degrees with both eyes.
The calculation of color difference is one of the most basic techniques in image processing fields. For example, color clustering and edge detection are the first steps of most image processes and we compute them by using a color difference formula. Although the CIELAB color difference formula is a commonly used one, the results obtained with it are not in accordance with human feelings when the color difference becomes large. In this paper, we have performed psychophysical experiments on color similarity between colors that have large color differences. We have then analyzed the results and found that the similarity is strongly restricted by the basic color categories. In accordance with this result, we propose a new color similarity metric based on the CIEDE2000 color difference formula and categorical color perception.
We propose a non-photorealistic rendering method for generating roller picture-like images from photographs. Roller picture-like images are artistic images painted using a roller. Our proposed method utilizes a modified bilateral filter (BF) called the weight-shifted BF. We validate the effectiveness of the proposed method by conducting experiments using the image of Lenna and other example images.