[in Japanese]

View full abstract

We experimentally fabricated a frame transfer CCD (FT-CCD) image sensor for Recommendation ITU-R BT, 601 with a single-layer poly-silicon electrode structure.We set the gap to 0.45 pm to balance charge-handling capability against optical sensitivity. We optimized the membrane structure to improve optical sensitivity. We sandwiched the poly-silicon gate between two silicon-nitride layers with a high refractive index to suppress reflection on the poly-silicon surface. The sensitivity was improved 40% with this structure. An FT-CCD image sensor constructed using this new structure is very simple and has high performance.

View full abstract

A wave optical simulator has been developed to analyze the optical characteristics of 3-D structures, including their diffraction and interference. The optical characteristics of inner-layer lens structures were studied by using this TOCCATA-Wave 3-D simulator. An inner-layer lens structure gives a better concentrated focal region and a smaller lateral shift in the focal region under oblique incident light compared with a single-lens structure. The improvement in the light-gathering power of the inner-layer lens structure was found to be remarkable when a small F-number and a large top-lens height were used.

View full abstract

CMOS image sensors with low power consumption and signal-processing circuits are widely used for mobile products and digital still cameras. A conventional MOS image sensor has high fixed pattern noise, up to the 30 mVp-p level, caused by the subthreshold voltage fluctuations of the cell-amplifier transistor and high random noises both the low-frequency cell-source follower circuits and the high-frequency output amplifier circuit. We developed and evaluated new circuit technologies to suppress both the fixed pattern noise in the pixels and the random noise in the cell-source follower and output amplifier. Using three kinds of analog circuit technologies, we developed an amplified 1/2-inch 1.3 M-pixel CMOS image sensor with low power consumption (60 mW) and high signal-to-noise ratio (50 dB).

View full abstract

We have developed a logarithmic-converting CMOS area image sensor that is fabricated using a general CMOS process. The sensor has 340 (H) ×228 (V) pixels, in which there is a logarithmic-converting circuit that uses MOSFET subthreshold operation. Its cell size is 20μm (H) ×20μm (V). It outputs a signal logarithmically proportional to the integrated amount of the photo current generated by incident light more than five orders of magnitude in intensity. It can therefore take bright and dark objects at the same time. We expect that this sensor will find consumer (i. e., digital still cameras, digital video cameras, and so on) and industrial application.

View full abstract

Abstract We have developed a CMOS area sensor for an auto-focus camera system, our AF SLR (Auto-Focus Single-Lens-Reflex) camera EOS-3 which achieves 45-point area auto focusing. It is fabricated using a 1.2 μm CMOS process technology. This sensor consists of new CMOS sensor cells, new CMOS memory cells, signal transfer circuits, and peak monitor circuits. This device realizes low noise characteristics by new noise cancellation technology. This sensor exhibits excellent performance, such as a high sensitivity of 26.6 V/lx·s and a wide dynamic range of 73.6 dB. It also has low power consumption, high-speed readout, and on-chip peripheral circuits.

View full abstract

An image sensor has been developed that consists of an X-Y-matrix field emitter array (FEA), a mesh electrode, and a high-gain avalanche rushing amorphous photoconductor (HARP) target, each of which is set closely facing one another. Investigation of its fundamental driving characteristics in a vacuum chamber showed that TV images are produced when the matrix FEA consists of 50×50 elements, that images are picked up with high sensitivity due to the HARP target, and that the output signal current is not affected by fluctuations in the emission current from the FEA under ordinary lighting conditions. Applying appropriate voltages to the mesh electrode improved the resolution. This image sensor will enable the development of super-high-sensitivity and ultrahigh-definition TV cameras.

View full abstract

Conventional image sensors, including CCD sensors and smart sensors are considered for just acquiring the image as the matrix of dots, not for recognizing the meaning of image.
In this paper, we propose a novel real-time algorithm to detect the square area in an object using the structure of node automata, and discuss its implementation as a CMOS image sensor, where the pixels and the node automata are integrated in one chip. We also discuss the search algorithm of their position in the pixel plain using area dividing methodology.

View full abstract

The depth information of objects is important for retrieving and indexing image information. A method is described for finding the depth of objects by taking three pictures, each with a different focus. One picture is focused at the target position, one is focused near the target, and the last one is far from it. DCT coefficients are used to estimate how each picture is focused. By comparing the sum of the frequency coefficients of these pictures, we can determine whether an object is in the target position or not.

View full abstract

We describe an image-registration method and a camera system for multi-focus image processing. We also describe a novel method for high-speed generation_ of arbitrarily focused images using graphic accelerator. The registration method uses global and local registration. The former corrects for differences in the displacement and scale between acquired images by using a hierarchical block-matching technique; the latter corrects for local deformation due to lens distortion. The developed camera can capture three differently focused images simultaneously for the same scene, so that, we can generate an arbitrarily focused image for dynamic scene. Finally, we show example motion images with an arbitrary focus generated from two differently focused image sequences captured by the proposed camera.

View full abstract

We describe a spectral-imaging system with a liquid-crystal tunable filter and algorithms for recovering the spectral functions of the illuminant and surface reflectance from the image data. The filter works as a polarization-interference filter with birefringent elements and linear polarizers ; the center wavelength of filtration is electronically tunable with no moving parts. The camera system is composed of a CCD camera, a personal computer, the filter, and its rotating stage. The system characteristics of (1) the influence of room temperature, (2) the spectral-sensitivity functions, and (3) the setting of filter axis are shown in detail. We define the camera outputs by a vector equation and describe the spectral functions by a linear model. Then algorithms are presented for estimating the spectral functions from the image data. An effective image-processing algorithm is proposed for highlight extraction and region segmentation. Experimental estimation of the illuminant and spectral reflectances, in which 21 spectral images were measured in the range of 450-650 nm, showed that the efficiency in measurement and the accuracy in estimation are better than those with the previous method.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

A HDTV plasma display with a large flat screen and high picture quality has been developed. The display uses a newly developed 42-inch diagonal DC-type PDP (Plasma Display Panel) with 1920 x 1035 display cells. The introduction of superior photolithography techniques for forming resistors, ribs, and priming ribs makes the fabrication of fine components more stable and reliable. And a new panel-driving method and signal-processing LSIs allow high-quality pictures to be reproduced on the panel. These technologies have opened the way for the development of a commercial 42-inch color PDP for HDTV.

View full abstract

Development of a Hi-Vision display system with a 42-inch-diagonal DC-type PDP is described. The configuration of the signal processing is discussed, and processing methods for the PDP are proposed. Development of signal-processing LSIs and ICs for the driving circuit is also described.
A pre-filter is needed for DC PDPs to avoid color aliasing. However, the sharpness of the displayed image is degraded by the filter. This degradation was compensated for by using an optimal-image enhancer. A resolution of 700 lines was thereby obtained. Processing to improve the gray-scale expression of displayed images was investigated. In some cases, an unnatural color was perceived in dark areas of the displayed images. To avoid this artifact, a pre-processing method for error-diffusion processing is proposed. Four signal-processing LSIs and two ICs for the driving circuit were developed. A high-quality 42-inch Hi-Vision PDP receiver, “Nagano Model”, was thereby achieved.

View full abstract

[in Japanese]

View full abstract