HSI Model-Based Image Dehazing for Remote Sensing Images

Abstract

Remote sensing images taken under hazy conditions have poor quality of the scene. Therefore, haze removal or dehazing is recommended. Due to the presence of haze, there is a resultant decay in the colour and the contrast of the captured image. Dehazing of an image is highly required to improve the quality of the scene. Algorithms based on dark channel prior combined with the haze imaging model were proposed by many researchers in recent years. Dark channel prior has been developed originally according to the statistics of outdoor haze-free images. When the dark channel prior is applied to remote sensing images, it often causes a colour drift phenomenon. This proposed work focuses on hue, saturation, and intensity colour model-based image dehazing, and this is a non-dark channel prior-based method. In this firstly, the atmospheric light is estimated from the hue of the hazy image as haze does not affect the hue. Then, the transmission medium is estimated from the saturation component. Finally, the haze is removed using two estimated parameters. The proposed work performs dehazing at a higher speed and makes it more sufficiently fast for a large-scale application which needs haze removal in the computer vision area. The experimental result shows that the proposed algorithm can achieve haze-free results while preserving the edges and it can improve the contrast of the image.