Haze affects the quality and visibility of the image. Many dehazing algorithms have been developed in recent years. However, the evaluation for the performance of the dehazing method is still not solved. The assessment is not easy to achieve since the reference image is not available. In this paper, a no reference image quality evaluation indicator is proposed to assess the visibility of a dehazed image. A multi-scale contrast feature is designed to measure the image sharpness. Considering some dehazing methods often cause under-dehazing results, a dark channel feature is employed to describe the haze residual degree of the restored image. Fusing the two features together, the final indicator that can measure the image visibility is obtained. Experimental results show that the assessment results are highly correlated with human visual perceptions and objective quality scores, which demonstrate the effectiveness and robustness of the proposed approach.
CITATION STYLE
Qin, M., Xie, F., & Jiang, Z. (2017). No reference assessment of image visibility for dehazing. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 10666 LNCS, pp. 664–674). Springer Verlag. https://doi.org/10.1007/978-3-319-71607-7_58
Mendeley helps you to discover research relevant for your work.