Certain investigation on visibility restoration and enhancement schemes for hazy and foggy images

Abstract

Current vision systems are designed to outperform in a pristine environment. Nowadays all the current vision systems are implemented with good quality sensors and hence the image obtained from such sensors is of very good quality. But the existing current vision systems fails to give good quality image when the sensors are not able to sense the reflectance from the object of interest, because of poor atmospheric conditions. The atmosphere modulates the reflectance from a scene point entering into the sensor due to the presence of atmospheric particles and reduces the clarity of the scene. This brings troubles to many vision-related applications such as Remote sensing, Surveillance, Military applications, Advanced Driver assistance systems, etc. by reducing the visibility of the scenes. newlineFactors affecting the visibility are haze, fog, smoke and cloud which degrade the quality of the captured images. To develop methods to enhance the visibility of the images degraded under weather conditions such as Haze, Fog and to find out the pollutants due to haze in the atmosphere. In order to implement this in real-time, various modules are developed by considering the computational complexity in this research. newlineThe first module deals with haze removal for remote sensing images. Dehazing of an image is highly required to improve the quality of the scene. Due to the presence of haze, there is a resultant decay in the color and the contrast of the captured image. Remote sensing images taken under hazy conditions have a poor quality of the scene. This proposed work, a non-Dark Channel Prior based method focuses on Hue, Saturation, and Intensity (HSI)color model-based image dehazing. From the results obtained, the proposed algorithm performs dehazing at a high speed and makes it more sufficiently fast for a large-scale applications that need haze removal in the computer vision area. newline