RCPID: retina color perception-based image dehazing

Abstract

Images captured from outdoor scenes are usually suffer from the atmospheric scattering and absorption, these natural phenomenon of hazy cause low contrast, poor quality of images, which severely affect their subsequent applications. Despite numerous image dehazing methods have been developed to improve the quality of hazy images, effective image dehazing method remains a challenging problem. In this paper, we propose a new method for retina color perception-based image dehazing, which considers the light response model based on the retina color perception mechanism, and the transmission estimation model combines with the dark channel prior. Specifically, an adaptive light response model is designed based on stimulus perception of retinal ganglion cells and lateral geniculate nucleus (LGN) neurones in the receptive field; and a new transmission estimation model combine with the dark channel prior is developed, thus we propose a retina color perception-based image dehazing method. Moreover, to validate the effectiveness of the proposed dehazing method, we test the method on various synthetic hazy images, and compare with several state-of-the-art image dehazing methods quantitatively. Extensive experimental results shown that the proposed method could improve the clarity of the hazy corrupted images, and the details of dehazed images could be protected moderately.