International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248 Volume: 4 Issue: 3 299 – 304 _______________________________________________________________________________________________ 299 IJFRCSCE | March 2018, Available @ http://www.ijfrcsce.org _______________________________________________________________________________________ Analysis of Fog Removal Technique Niharika Singh 1 , Ashima Arora 2 , Sahil Kasana 3 , Sangeeta Dhall 4 , Shailender Gupta 5 YMCA University Of Science & Technology, Faridabad niharika.3297@gmail.com, ashimaarora95@gmail.com, sahilkasana546.sk@gmail.com, sangeeta_dhall@yahoo.co.in, shailender81@gmail.com Abstract—In simple terms, fog is an atmospheric phenomenon which creates a sense of confusion and complexity to vision. In presence of fog, perception drops down to less than 1000 m. This happens due to plenty of reasons such as bad weather, haze, mist, smoke etc. In economic terms as well, it is not a positive phenomenon— transportation system is affected severely, so is aviation, navigation, surveillance. Hence it becomes imperative to devise fog reduction techniques. Over the years, researchers have come up with various techniques to reduce fog. Effectively, the composition of fog consists of air-light and direct attenuation. Air-light is because of scattering of light due to water droplets in the air which happen to make the scene appear whiter than normal, and attenuation is gradual decay in intensity of flux through a medium. This paper focuses on literature survey of techniques such as Dark Channel Prior (DCP), Improved Dark Channel Prior (IDCP), Anisotropic Diffusion, DCP with histogram specification, Improved DCP Using Guided Filter. The techniques discussed in this paper lay the foundation of results based on the following parameters: Normalized Colour Difference (NCD), Contrast Gain (C Gain), Number of Saturated Pixel, Colour Naturalness Index (CNI), Time Complexity (TC), Perceptual Quality (PQ). The software used for evaluation of efficacy is MATLAB-2015. It is observed that the best perceptual quality is obtained for IDCP with Guided Filter followed by IDCP, DCP with Histogram Specification, Anisotropic Diffusion and DCP. Keywords- Anisotropic Diffusion, Histogram, Attenuation, Pixel __________________________________________________*****_________________________________________________ I. INTRODUCTION Fog [1-2, 20-21, 23] tends to get formed when water is suspended in the air just like cloud but at ground level. With increase in pollution, the thickness of fog increases because the particles in air allow more water droplets to get condensed. There are different kinds of fogs occurring in nature. The categories are made based on the process that causes water droplets to form in the air. For example; radiation fog, freezing fog, valley fog, evaporation fog, advection fog. Radiation fog happens to be a seasonal phenomenon, mostly in winter. It disappears when the sun rises. It is caused by cooling of land overnight and the thermal radiation then cooling the air close to the surface. Condensation of water content occurs when air is no longer able to hold its moisture. Freezing fog occurs when water droplets remain in liquid state even in spite of temperature falling below freezing point. The condition of valley fog may go on for days. The concept behind valley fog is that when dense, cold air settles at the bottom of a valley and hotter air passes above the valley. Evaporation fog is a local phenomenon which leads to formation of frost. It happens when cold air passes over warm water and moist land. This results in formation of mist. This is a common sight around hot tubs. Advection fog is habitually seen around coastal areas. This occurs when air with high moisture content passes over a cool surface. Certainly, condition of fog is a serious weather condition. Fog has a lot of direct consequences on people‘s everyday life- health, aviation, road transport, surveillance, tracking, loss in basic visibility etc. The main reason for loss in basic visibility is scattering of light in the air due to presence of condensed water droplets. Fog is made up of two major constituents: air-light [1, 6] and direct attenuation [1, 6]. Air-light is an effect caused because of scattering of light. Air-light makes the scene appear whiter which results in reduction in image quality [3]. Mathematically, the equation of air-light is given below (see equation 1) Air-light = A (1-e -βd(x) ) … (1) Where, A is global atmospheric light β is scattering coefficient of atmosphere d(x) is scene depth of xth pixel Attenuation, in very simple language is a general term that refers to gradual reduction in strength of a signal through a medium. It sometimes called ‗loss‘. Direct attenuation illustrates scene radiance and its decay in the medium. Mathematical expression of direct attenuation with respect to fog is [1-2, 5-6], Direct Attenuation = J(x). T(x) … (2) Where, J(x) is scene radiance T(x) is medium transmission When the atmosphere is homogeneous, medium transmission T(x) can be expressed as T(x) = e -βd(x) … (3) This means scene radiance is attenuated exponentially with depth. Fog = Air-light + Direct Attenuation … (4) Or in mathematic form, fog can be rewritten as: Fog = J(x). T(x) + A (1-t(x)) … (5) Organization of remaining paper is as follows; section two throws light on literature survey of five fog removal techniques[7-14, 21-22] namely: Dark Channel Prior (DCP) method [7-8], Improved Dark Channel Prior (IDCP)[11], IDCP with histogram specification[14], Anisotropic Diffusion[15], IDCP with guided filter[12-13]. Section three summarizes simulation setup parameters which contains setup parameters and performance metrics used. The fourth section compares results of all techniques. Finally, fifth section concludes the survey by comparing and summarizing all the results obtained. II. LITERATURE REVIEW Present section focuses on all five fog removal techniques previously mentioned. These techniques use a single image at a time.