http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
신태민,박규태,박구만,강현철,최동혁 연세대학교 산업기술연구소 1987 논문집 Vol.19 No.2
In this paper a parallel processing system using SIMD architecture is implemented for image processing. The system consists of a set of 4 PEs(Processing Elements), a CU(Control Unit), an ADG(Address Generation Unit), and an interconnection. A CU is implemented by the microprogram and controls PEs and ADG. Microprogramed CU enables to process in high speed and upgrade the flexibility of the system. For efficient addressing, next address is generated at ADG and broadcast to each PE using pipeline technique. To implement a simple interconnection, the time sharing common bus is used. The characteristics of this proposed system are ease of implementation, high speed, flexibility of the system and high expandibility of PE. To evaluate the perfomance of the system, window processing is used. The number of the expandable PEs, execution time, speed up, and efficiency are evaluated.
박규태,전중남,신태민,연창모 연세대학교 산업기술연구소 1987 논문집 Vol.19 No.2
This study describes an interframe image coding method using motion compensated and classified vector quantization(MC-CVQ). In motion compensated vector quantization(MCVQ), it is very important to encode the significant block. Therefore, we propose the MC-CVQ which encodes the significant block in the motion compensated error image by the classified vector quantizer. We classify the significant blocks according to the positions of the significant vectors, and also classify the significant vectors into 15 classes according to the positions of the significant pels. The computer simulation shows that the signal-to-noise ratio and the average bits of MC-CVQ are 0.2-0.25 bits/pel and 35 to 37 dB, respectively.