Implementation of the High-Speed Feature Extraction Algorithm Based on Energy Efficient Threshold Value Selection

• Published in: Transactions on electrical and electronic materials 2020-04, Vol.21 (2), p.150-156
• Main Authors: Lee, Juseong, An, Ho-Myoung, Kim, Jooyeon
• Format: Article
• Language: English
• Subjects: Chemistry and Materials Science; Electronics and Microelectronics; Instrumentation; Materials Science; Optical and Electronic Materials; Review Paper
• ISSN: 1229-7607; 2092-7592
• DOI: 10.1007/s42341-020-00188-x

This paper proposes an implementation method of high-speed feature extraction algorithm. The proposed approach is based on the block type classification algorithm, which can reduce the computational complexity required for feature detection by re-using the threshold value when the same block type occurred in same index of successive image. The experiment was carried out for quantitative analysis as the image resolution and block size were changed. The experiment results have shown that the better data consistency with the larger the block size and the smaller the image resolution, the matching probability of the threshold value was 97.99% on average when the block size was 128 × 128. Applying the proposed method to the Canny edge detection can completely eliminate the latency required for image processing of adaptive threshold value calculation, which requires significant computing complexity if the block type overlaps with the previous frame. By applying block type sorting algorithm to various feature detection algorithms in this way, it is expected that time needed for computation can be reduced.