September 22, 2006, Kawasaki, Japan--Fujitsu Laboratories Limited has unveiled image-compression technology for automotive use, designed to be paired with the high-capacity multiple video channels being developed for cars. Such systems (such as this one) will handle video from numerous cameras and other inputs on the inside and outside of a car; the inputs could be from night-vision systems, IR imagers, visible imagers used as rearview-mirror replacements, DVD players for passengers, GPS, and so on.
Fujitsu's technology, called SmartCODEC, complies with standards for in-vehicle digital-imaging transmission-system networks. It has a compression rate of one-third and a compression-decompression time of a mere 2 to 3 milliseconds. As well as making higher-capacity video possible, the technology reduces the amount of cable required within the vehicle to achieve rapid, high-quality transmission of multiple images.
The analog signals that have been used thus far require that a separate cable be wired for each channel, a process that has become complicated as the number of channels increases. This has driven a switch to systems capable of multiplex transmission of images on a single in-vehicle local-area network.
Even if bandwidth meets the IDB-1394 image-transmission standard (transmission speeds of 400 Mbit/s), it is insufficient for uncompressed transmission of multiple images. Developing image-compression technology for automotive use involves overcoming three challenges: making low-cost packaging for each compression and decompression channel; achieving a transmission delay time of less than 30 milliseconds for surveillance cameras used when the vehicle is being driven; and achieving clear reproduction of letters and lines on images transmitted for car-navigation systems. In image-compression methods used thus far, such as MPEG, external memory was required for compression-decompression of large-scale integrated circuits, leading to higher cost. In addition, the compression-decompression delay time of several hundred milliseconds had to be overcome. Conventional methods mainly used natural images, and there were concerns about poor image quality for car navigation systems, such as broken or indistinct letters.
The SmartCODEC was developed to be light in weight and have a minimal delay time and high image quality. In order to lighten the circuitry, a differential-pulse-code modulation scheme featuring simple structures was adopted. To clearly reproduce camera images and the edges of letters and lines on the navigation system, the modulation level was optimized at the pixel unit. By encoding pixels in groups of two, the processing speed was accelerated without increasing the basic clock speed, enabling minimal delay time.
Target images handled by the technology include 720 x 480 pixel YUV images (which are based on luminance [Y] and chroma [U and V] values) and 800 x 480 pixel RGB (red-green-blue) images. SmartCODEC can process the latter images at a maximum of 60 frames per second. The transmission rate for 30 frames per second is 55 Mbit/s.
The technology is scheduled to be commercialized by Fujitsu in early 2007.