This paper proposes a new algorithm of a full-video object search method which extracts every video frame in which a specified object appears. The new algorithm aims principally at reduction of oversights in the object detections. It devides a specified object image into some characteristic regions, chooses characteristic colors in each region and makes a group of colors as a feature of each region. It is decided that a specified object appears in a frame containing sufficient pixels which represent colors in each group. To make the detections precise, relative locations of each color regions are also examined.

View full abstract

The color of reflection objects is changed by illuminants. The color gamut and the color difference of color photograph caused by illuminant changes in CIE 1976 L*a*b* color space are evaluated. They are calculated by illuminants A, F6, F2, D50, F8, D65 and C. As the results of evaluation, illuminants D50 and F8, both color temperature are 5000K, have the largest color gamut of all the meassured illuminants and reproduce the most colorfulness. Moreover, these illuminants are nearly equivalent for color evaluation.

View full abstract

Although JPEG algorithm is the standard of color still image coding method, color space is not specified. Quality evaluation using various color space, such as NTSC-RGB, CIEXYZ, CIELAB, CIELUV, YCbCr is described. The relation between color space and image quality with JPEG coding is analyzed. By objective evaluation, CIELUV is the best, and CIELAB, YCbCr, RGB, XYZ follow, and by subjective evaluation. YCbCr is the best, and CIELUV, CIELAB, RGB, XYZ follow, although in both cases, the difference of the results of top three is not so significant. We developed subjective assessment and analysis methods, and these will be useful for future image quality evaluation.

View full abstract

we propose new techniques for high quality facsimi le document reading. First one is Saddle Filter. Saddle Filter is an edge enhancing filter which supresses Moire pattern of halftone image and Notch pattern of binarized image. Second one is Spreaded Dot Correction. Spreaded Dot Correction improves the rendition of the tonal scale of the original image. We have implemented these techniques on a LSI, named Facsimile Image Processor FRIP2 (MN86072). In this paper, we present the algorism of these techniques and exhibit the experimental results. We also show the specifications of the FRIP2 (MH86072).

View full abstract

We have developed a 32-inch high-definition television with a full specification MUSE decoder. A high fidelity HDTV waveform was restored with 10-bit phase equalization circuit in this decoder. A field doubling system was used for current NTSC signal processing to realize same picture quality as a broadcasting monitor. A MUSE-NTSC converter was installed to enable recording to current VCR Features like zoom- / full-mode picture, PinP / PoutP, channel index were implemented. Picture control, recording reservation, etc were manipulated using high-resolution graphic panel.

View full abstract

SONY has put a new full-spec Hi-Difinition TV set on the market. The new TV has the 2nd generation MUSE decoder system. FUJITU. HITACHI, Texas Instruments Japan and SONY cooperate for the system, which includes newly developed 13 kinds of LSIs. We reduced the number of LSIs to 30, one-fifth of 1st generation decoder system. We realized the reduction of LSIs and less power consumption by applying the most advanced LSI process, which is about 0.8 to 1μm. And we also produced LSIs for I/O and peripheral circuits. It results in less power consumption and less adjustment process in a factory. we contained low transmissive destortion (less than 0.2%) by using A/D converter with 10 bit/32.4MHz and apllying a new waveform equalization algorithm.

View full abstract