Compressed-image recognition for a feature extractable CMOS image sensor
Abstract
To realize an energy-efficient image recognition system, we propose a CMOS image sensor that directly outputs compressed feature data from the captured image. This sensor outputs binarized luminance signals and edge signals as the feature data, significantly reducing the data size with run-length encoding (RLE). Simulation results show that, compared to conventional RGB color images, the proposed sensor reduces the output data size by 98.8% and achieves an image recognition accuracy of 84.4% without requiring feature data decoding.
Fields of Study & Subjects
CCD and CMOS Imaging Sensors
Advanced Neural Network Applications
Advanced Data Compression Techniques
Publication details
- Journal / source
- International Workshop on Advanced Imaging Technology (IWAIT)
- Pages
- 60–60
- Publication date
- 27 February 2026
- Language
- EN