Embedded lowlight image quality measure

Abstract

Based on the research on embedded camera user's attention level, a novel blind image quality ranking and poor appearance detection algorithm for low lighting environment are proposed for the first time. Different from the traditional algorithms, the proposed one mixed model composed of detecting lowlight, grading blur degree and noise level, and figuring out the under- or over-exposure. It reaches the real time performance even on low level embedded devices with the detail feedback of quality score. The experiments show the proposed method performs well and is consistent with human perception.