Title: Underwater images quality assessment and its related enhancement algorithm

Speaker: Prof. Pengfei Guo （Zhongkai University of Agriculture Engineering）

Time: Fri, Sep.18 2020, AM:09:00-10:00

Location:  4131, Building No.4, Wushan Campus

Inviter: Prof. Zhujun Zheng

Abstract:

        The enhancement of the underwater culture image is ana impor- tant problem for the engineering of the marine industry. In this paper, we study 5 popular underwater culture image enhance- ment alogorithms- contrast limited adaptive histogram equalization enhancement(CLAHE), decorrelation stretch enhancement(DS), underwater dark channel prior enhancement(UDCP), non-local image dehazing(NLID) and enhancement underwater image by fusion(EUIF). We apply these algorithms to enhance the various images under different reality culture environment. In order to assessment the performances and adjustment of these algorithms for closing the human visual system, we have created a images dataset with several sea cucumber images under various reality underwater culture environment. We analysis the histogram of mean subtracted contrast normalized coefficients of original images and the enhanced versions by the above 5 algorithm. The image visual quality provided by these algorithms is evaluated subjectively. We also proposed two novel non- linear no- reference objective underwater quality metric based on the results of above subjective qualification. As an application, we poporsed a multiscale fusion enhancement algorithm for underwater images based on the quality inspired weighted maps.