In this paper, we proposed SIMT based Image Signal Processor which can apply various image preprocessing algorithms and allow parallel processing of application programs such as image recognition. Conventional ISP has the hard-wired image enhancement algorithm of which the processing speed is fast, but there was difficult to optimize performance depending on various image processing algorithms. The proposed ISP improved the processing time applying SIMT architecture and processed a variety of image processing algorithms as an instruction based processor. We used Xilinx Virtex-7 board and the processing time compared to cell multicore processor, ARM Cortex-A9, ARM Cortex-A15 was reduced by about 71 percent, 63 percent and 33 percent, respectively.

Abstract. This paper proposes implementation of CENTRIST-based pedestrian detection in embedded environments. Although a considerable number of pedestrian detection algorithms have been proposed, they are not suitable for implementation in embedded environments. In this paper proposes a CENTRIST-based pedestrian detection method which combines census transform and histogram, instead of the HOG method, which is more complicated. An Aldebaran board (300MHz) was used for implemntation of the algorithm in embedded enviornments, and the 512x360 pixel input image was used. The implemented algorithm demonstrated performance of 0.7 frame per second.

No abstract

Abstract. In this study, SIMT based ISP possible for processing various image pre-processing algorithms and recognition algorithms was proposed. The conventional ISP is hard-wired designed to process faster, but it has limits in performing various image processing algorithms. The proposed ISP is the command-based processor performing various image processing and improves the processing speed. For verification, we implemented 2-Radix FFT operation and pedestrian recognition using the proposed ISP. The study results showed approximately 2.7 times improved processing speed than the embedded platform CPU of Cortex-A9, and about 1.5 times improved than the Cortex-A15. Keywords:We would like to encourage you to list your keywords in this section.

No abstract