Development of High Performance Hardware by High-Level Synthesis of Median-Based Dynamic Background Subtraction Method with Multiple Line Buffers

被引:0
作者
Shinyamada, Kohei [1 ]
Yamawaki, Akira [1 ]
机构
[1] Kyushu Inst Technol, Dept Engn, Kitakyushu, Fukuoka, Japan
来源
2021 IEEE REGION 10 CONFERENCE (TENCON 2021) | 2021年
关键词
FPGA; Image Processing; HLS;
D O I
10.1109/TENCON54134.2021.9707206
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Hardware processing is suitable for embedded image processing systems because of its lower power consumption and higher performance compared to software processing. To facilitate development, a tool called high-level synthesis, which automatically converts high-level languages into hardware description languages, is used. However, high-level synthesis of pure software does not necessarily generate efficient hardware. In this study, we attempted to generate high-performance image processing hardware using a median-based dynamic background subtraction method. As a result, we found that high-performance hardware can be generated when multiple line buffers are introduced. Compared to the non-introduced one, the performance was improved by about 13 times.
引用
收藏
页码:123 / 127
页数:5
相关论文
共 7 条
  • [1] [Anonymous], 2016, THE NEXT PLATFORM
  • [2] [Anonymous], 2016, BERTEN 0519
  • [3] Calkins H, 2017, J ARRYTHM, V33, P369, DOI 10.1016/j.joa.2017.08.001
  • [4] Detecting moving objects, ghosts, and shadows in video streams
    Cucchiara, R
    Grana, C
    Piccardi, M
    Prati, A
    [J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2003, 25 (10) : 1337 - 1342
  • [5] Ishikawa Y., 2017, Proceedings of the 8th International Symposium on Highly Efficient Accelerators and Reconfigurable Technologies, P1
  • [6] jp mathworks, HIGH LEVEL SYNTHESIS
  • [7] Shinya UIDA, 2019, J I IMAGE ELECT ENG, V48, P237