Unsupervised and optimized thermal image quality enhancement and visual surveillance applications

Thermal images suffer from low-luminance issues under specific conditions, such as heat radiation, distance-to-radiated objects, reflection angles. Low-luminance thermal images degrade the performance of a vision-based system. Present thermal contrast enhancement algorithms usually tolerate the background, noise amplification, and brightness distortion from over-enhancement. This paper introduces; i) an image enhancement algorithm for thermal satellite and aerial images; ii) a novel thermal image quality assessment; iii) a modified nature-inspired algorithm to search optimal parameter for the proposed enhancement algorithm; and iv) the application of the photovoltaic panel inspection by using the proposed algorithms. Experimental results thermal photovoltaic panel images with various luminance scenes from two datasets demonstrate the proposed method’s effectiveness compared with state-of-the-art algorithms both qualitatively and quantitatively. Moreover, the proposed algorithm performs favorably for low-luminance thermal images with different illumination levels while enhancing the dark regions’ details and preserving mean-brightness.