Haze Simulation Based on a Physical Modeling and Improved Image Visibility Restoration

Abstract

In this paper, we design hazing based on a physical modeling and propose a method to dehaze by efficiently estimating the transmission amount from fog images. He [9] proposed a single image visibility restoration method using dark channel prior (DCP). Among many algorithms, this DCP algorithm is known to have good performance in various fog images. However, this method requires matting processing, which requires a large amount of computation to refine the block type of the transmission amount. In this paper, we propose an improved image visibility restoration algorithm that estimates the transmission amount using a hybrid-type multi-filter based on edge intensity information to more accurately remove haze at the boundary. Through the experiment of a proposed algorithm, we show that haze is effectively eliminated without using matting processing. In particular, the algorithm is more effective at the boundary region.