Low Complexity Single Color Image Dehazing Technique

Low Complexity Single Color Image Dehazing Technique

SANGITA ROY (NARULA INSTITUTE OF TECHNOLOGY, India) and Sheli Sinha Chaudhuri (Jadavpur University, India)
Copyright: © 2018 |Pages: 14
DOI: 10.4018/978-1-5225-5246-8.ch004

Abstract

Atmospheric particulate matter (APM) disturbs the biotic environment creating global warming, health hazards, and last but not the least, visibility reduction. Removing haze from degraded image is an extremely ill-posed inverse problem. Real time is another factor influencing the quality of resulting image. There is a trade-off between time and quality of resulting image. Single image makes the algorithms more challenging. In this work, the authors have developed fast, efficient, single gray or color image visibility improvement algorithms based on image formation optical model with minimum order statistics filter (MOSF) as the transmission model. The quantitative and qualitative evaluations of the proposed algorithms have been studied with other existing algorithms. The result shows improvements over existing state-of-the art algorithms with minimum time. Time has been evaluated by execution time along with time complexity Big (O). The resultant images are visibly clear satisfying the criteria for computer vision applications.
Chapter Preview
Top

Introduction

APM or aerosol is a real threat in our green world. Researchers are fighting against it. One of the reasons of APM is unplanned civilization and technological advancements. It has been observed from satellite images that Asian and African and very few parts of American countries are the most polluted places. Image sources from satellites show year wise degradation .APM are both natural and manmade. APM are a mixture of solid and liquid droplets. They are in variety of sizes. Coarse APM ranges PM10-PM2.5 (micrometre diameter), finer are below PM2.5 and ultrafine below PM0.1(Mao, 2015). Computer vision (CV) is a promising branch of technology which encompasses object tracking, object recognition, surveillance, image enhancement etc. Clear image is an essential requirement in CV. But APM degrades visibility of the received image. Rain, fog, vog, mist, fume, smog, hail, snow etc. are the cause of visibility reduction in image. Outdoor image is a challenge in computer vision, especially in bad weather condition. Image formation at the viewer point (i.e. may be considered camera) is influenced by distance, airlight, transmission, and scattering coefficient [Koschmiede, 1924). There are some classical enhancement techniques, like, histogram equalization, imadjust, adaptive histogram equalization. These techniques work well in most of the cases. In some special cases (like, fog, haze, smoke, rain, vog etc.) these techniques fail. In those special cases of bad weather nothing can be seen from the image. The situation becomes worsen if no ground truth or reference image could be found. In these contexts special popular algorithms have to be selected. Those are working with no ground truth image or popularly called single image dehazing algorithms. The key observations found from outdoor haze free images are as i) image contrast of normal image is high, ii) airlight does not affect the richness of the image, iii) pixels intensity is well distributed in the intensity scale, and iv) pixel over-saturation and under saturation do not exist(Xhang, et al.,2004]. Contrary to that of hazy images are of low contrast and airlight makes images white. Most of the pixel intensities are very high i.e., under-saturated and flocked together. Degraded image pixels are over saturated in one of the channels. The cause may be due to the illuminant of strong colour cast or sensor/camera respond differently for different colour channels. Resultant of these artifacts makes image pixels achromatic. Visibility Improvement is under the category of Ill-Posed Inverse Problem. In this class best or optimum image has to be extracted from a series of attenuated received images. Sometimes it is extremely difficult to retrieve any information about the original image as the problem becomes Ill-Posed Inverse Problem (Roy, Chaudhuri, 2016). Paper is arranged as below. Section II consists of Literature survey. In section III Minimum OSF detail study has been presented. Proposed methods have been illustrated Section IV. Result is described in section V. Section VI explains qualitative and quantitative analysis. Section VII elaborates its applications and type of device used for the research. Finally section VIII is for Conclusion.

Complete Chapter List

Search this Book:
Reset