IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308 __________________________________________________________________________________________________ Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org 432 IMPROVED SINGLE IMAGE DEHAZING BY FUSION Nitish Gundawar 1 , V. B. Baru 2 1 ME Student, Department of Electronics and Telecommunication, Sinhgad College of Engineering, Pune, 2 Associate Professor, Department of Electronics and Telecommunication, Sinhgad College of Engineering, Pune, Abstract One of the major problems in image processing is the restoration of images corrupted by various types of degradations. Images of outdoor scenes often contain atmospheric degradation, such as haze and fog caused by particles in the atmospheric medium absorbing and scattering light as it travels to the observer. Although, this effect may be desirable from an artistic stand point, for a variety of reasons one may need to restore an image corrupted by these effects, a process generally referred to as haze removal. This paper introduces improved haze removal technique based on fusion strategy that combines two derived images from original image. These images can be obtain by performing white balancing and contrast enhancement operation. These derived images are weighted by specific weight map followed by Laplacian and Gaussian pyramid representations to reduce artifacts introduce due to weight maps. Unlike other techniques this approach requires only original degraded image to remove haze which makes it simple, straightforward and effective. Keywords: Outdoor applications, fusion, dehazing, image pyramid -----------------------------------------------------------------------***---------------------------------------------------------------------- 1. INTRODUCTION Images of outdoor scenes often contain haze, fog, or other types of atmospheric degradation caused by particles in the atmospheric medium absorbing and scattering light as it travels from the source to the observer. Image obtained at other end is characterized by reduced contrast and faded colours. While this effect may be desirable in an artistic setting, it is sometimes necessary to undo this degradation. Weather conditions differ mainly in the types and sizes of the particles involved and their concentration in space. A great deal of effort has gone into measuring particle sizes and concentrations for a variety of conditions as shown in table I. For example, many computer vision algorithms rely on the assumption that the input image is exactly the scene radiance, i.e. there is no disturbance from haze. When this assumption is violated, algorithmic errors can be catastrophic. One could easily see how a car navigation system that did not take this effect into account could have dangerous consequences. Accordingly, finding effective methods for haze removal is an ongoing area of interest in the image processing and computer vision fields. This task is important in several outdoor applications such as remote sensing, intelligent vehicles, underwater imaging and many more. In this paper improved fusion based haze removal technique is discussed. The main concept of fusion is to combine two or more images into single image that can be more suitable for some intended purposed [16]. Therefore, image fusion is effective technique that is designed to maximize relevant information into fused image. Table-1: Weather conditions and associated particles types, sizes and concentration [2] Conditions Particle Size Radius (μm) Concentration (cm −3 ) Air Molecule 10 −4 10 −19 Haze Aerosol 10 −2 - 1 10 3 - 10 Fog Water Droplet 1 - 10 100 – 10 Cloud Water Droplet 1 - 10 300 – 10 Rain Water Drop 10 2 −10 4 10 −2 - 10 −5 The main idea behind fusion based dehazing technique is to combine images derived from degrade image. Two images are derived by performing white balance and contrast enhancement operation on original degraded image. This ensures the visibility in hazy and haze free region of image and also eliminate unrealistic color cast introduced due to atmospheric color. In fusion framework the derived inputs are weighted by three weight maps i.e. luminance, chromatic and saliency weight maps [1]. These weight maps ensure to preserve regions with good visibility. However, artifacts introduced by weight maps can be eliminated by fusing Laplacian pyramid representation of derived inputs and Gaussian pyramid representation of normalized weight that yields dehaze version of original degraded image. The rest of the paper is structured as follows. Below in section 2 previous dehazing methods are briefly discussed. In section 3 theoretical aspects of light propagation is discussed. In