BROWSE
Detail
Conference Paper
Embedded Census Transform을 적용한 실시간 얼굴 검출 하드웨어 구조의 제안
Cited - time in 
Share 
Share
- Issue Date
- 2011-06
- Citation
- 대한전자공학회 종합 학술 대회 (하계) 2011, pp.631-634
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- In this paper we propose the hardware architecture for memory optimization of Altair chip which is a real-time face detection ASIC. Altair can detect 32 frontal faces a frame per second on QVGA(320x240) color image from CMOS camera. For solving the problems of heavy image processing, it was designed to have multi frame buffer and parallel LUT memory configuration then a great part of chip is internal memory area. Consequently, we introduce a modification to MCT feature transform to reduce internal usage of memory in Altair. The new method(Embedded Census Transform) is cascade structure of bilateral filter and census transform with bias. We confirm that internal memory usage decreases by 30% after applying ECT to Altair hardware architecture.
- KSP Keywords
- Bilateral Filter, CMOS camera, Cascade structure, Census Transform, Color images, Frame buffer, Image processing(IP), Internal usage, Memory Configuration, Memory optimization, Real-Time Face Detection