Image dehazing using Gaussian and Laplacian Pyramid

Transcription

1 Image dehazing using Gaussian and Laplacian Pyramid 1 Chhamman Sahu, 2 Raj Kumar Sahu Dept. of ECE, Chhatrapati Shivaji Institute of Technology Durg, Chhattisgarh, India chhammansahu007@gmail.com, rajkumarsao@gmail.com Abstract Image processing techniques improve the quality of an image and enhance the maximum information from the degraded image. The proposed method adjusts the global contrast in the spatial domain for dehazing fog or haze spread over images and then enhances the local contrast in the transform domain for reviving the details of images. It is the process of combining two or more images into a single image; method is a fusion-based strategy that derives from two original hazy images inputs by applying a white balance and contrast enhancing procedure. The method performs in a pyramidal method, which is straightforward to implement. Then the resulting image will be more clear and enhanced from the prior. I. INTRODUCTION Image are degraded by the bad weather condition like fogs, ash volcanic ash, dust, smoke all these particle are provide obstacle in clearness of images. The quality of photograph in our daily life is easily undermined by the aerosols suspended in the medium, such as dust, mist, or fumes. This has an effect on the image, e.g., contrasts are reduced and the surface color becomes fad. Such degraded photographs often lack visual vividness and offer a poor visibility of the scene contents. The goal of haze removal algorithms is to enhance and recover the detail of the scene from haze image. There are many circumstances that accurate haze removal algorithms are needed. In computer vision, most automatic systems for surveillance, intelligent vehicles, object recognition, etc., assume that the input images have clear visibility. However, this is not always true in bad weather. In consumer photography, the presence of fog will be an annoyance to the images for it reduces the contrast significantly.image fusion is the good information from each of the given images is fused together to form a resultant image whose quality is superior to any of the input images. This is achieved by applying a sequence of operators on the images that would make the good information in each of the image prominent. II. LITERATURE REVIEW In [1], Codruta proposed an algorithm based on independent component analysis. This algorithm estimates the surface shading and the medium under the assumption transmissions which are locally uncorrelated, and gets the image local albedo and restore image contrast. However, a distinct lack of enough color in heavily haze images makes heavily haze images cannot be handled well. Tan [11] developed a system for estimating depth from a single organization). This template was degraded input image. Motivated by the fact that contrast is reduced in a foggy image, Tan divided the image into a series of small patches and postulated that the corresponding patch in image should have a higher contrast (where contrast was quantified as the sum of local image gradients). Tan [11] made the observation that a haze-free image has higher contrast than a hazy image, and was able to obtain good results by maximizing contrast in local regions of the input image. However, the final results obtained by this method are not based on a physical model and are often unnatural looking due to over-saturation. Fattal [12] assumed every patch has uniform reflectance, and that the appearance of the pixels within the patch can be expressed in terms of shading and transmission. He considered the shading and transmission signals to be unrelated and used independent component analysis to estimate the appearance of each patch. Significant progress in single image haze removal has been made in recent years. In [13], He proposed a simple but effective image prior-dark channel to remove haze from a single input image. However, as the haze imaging model assumes common transmission for all color channels, this method may fail to recover the haze images. In [14], Tarel proposed an improved image defogging algorithm based on bilateral filtering. Fattal [12] was able to obtain good results by assuming that transmission and surface shading are locally uncorrelated. With this assumption, he obtains the transmission map through independent component analysis. This is a physically reasonable approach, but this method has trouble with very hazy regions where the different components are difficult to resolve. Lastly, a simple but powerful approach proposed by He [13] uses dark pixels in local ISSN (Print): , Volume-3, Issue-4,

2 windows to obtain a coarse estimate of the transmission map followed by a refinement step using an image matting technique. Their method obtains results on par with or exceeding other state-of-the-art algorithms, and is even successful with very hazy scenes. A. Theoretical level White balancing is an important processing step that aims to enhance the image appearance by discarding unwanted color casts caused by the atmospheric color. Due to the fact that haze is dominating the image, an average value is computed for the entire image. Similar as in [1], a straightforward biasing of the image average color towards pure white is employed. This step assures that atmospheric light color constant is equal to one and the normalized image values are in the range [0, 1]. As observed in [1], when the light color varies in the image it is more robust to perform this bias operation using the local average value, as shown in Fig. 2b. Practically, the first input of the fusion process is computed based on the straightforward white balancing operation. Nevertheless, white balancing solely is not able to solve the problem of visibility, and therefore an additional input is needed to enhance the contrast of the degraded image. some generic requirements can be imposed on the fusion result: a) The fusion process should preserve all relevant information of the input imagery in the composite image (pattern conservation) C. Fusion Process The images are first decomposed using a Laplacian Pyramid decomposition of the original image into a Hierarchy of images such that each level corresponds to a different band of image frequencies [1]. B. Weights of the Fusion The design of the weight measures needs to consider the desired appearance of the restored output. Since image restoration is tightly correlated with the color appearance, so the measurable values such as haze density, salient features and exposedness are difficult to integrate by naïve per pixel blending. Higher values of the weight determine that a pixel is advantaged to appear in the final image. Fig.2. Flowchart of dehazing the image of foggy images Fig.1. Pixel Pyramid Level 1) Pixel Level Fusion This section focuses on the so-called pixel level fusion process, where a composite image has to be built of several input images[1]. In pixel-level image fusion, The Laplacian pyramid decomposition is a suitable MR decomposition for the present task as it is simple, efficient and better mirrors the multiple scales of processing in the HVS. The next step is to compute the Gaussian pyramid of the weight map. Blending is then carried out for each level separately [1] [2]. 30

3 Fig.3. Many Haze Images and Dehazed Image After Removing the Haziness 1) Inputs In practice, there is no enhancing approach that is able to remove entirely the haze effects of such degraded inputs. Therefore, considering the constraints stated before, since we process only one captured image of the scene, the algorithm from the original image only two inputs that recover color and visibility of the entire image. The first one better depicts the haze-free regions while the second derived input increases 2) Weight Map As can be seen in figures by applying only these enhancing operations, the derived inputs still suffer visible details of the hazy regions. Inherently inspired by the previous dehazing approaches such as Tan [11], Tarel and Hautiere [14] and He et al. [13],we searched for a robust technique that will properly white. Table I. Calculation For PSNR,ERMS and Gray values PSNR ERMS GRAY VALUE Image Image Image Image Image Image Image Table II. Entropy for Haze and Dehazed Image Entropy of Haze images Entropy of Dehazed images Image Image Image Image Image Image Image

4 D. Result and discussion Prove the robustness of our method, the new operator has been tested on a large dataset of different natural hazy images. Haze due to dust, smoke and other dry particles reduces visibility for distant regions by causing a distinctive gray hue in the captured images. However, our technique has been successfully tested as well for a slightly different case: foggy scenes (e.g. the first and the third example in figure 6; the reader is referred also to the supplementary material for more cases). For our problem, fog has a similar impact as haze, but technically it appears as a dense cloud of water droplets close to the ground when night conditions are clear but cold, and the heat released by the ground is absorbed during the day (please refer to figure 12). We have shown that, by choosing appropriate weight maps and inputs, a multi-scale fusion strategy can be used to effectively dehaze images. Our technique has been tested on a large data set of natural hazy images. The method is faster than existing single image dehazing strategies and yields accurate results. In future work we would like to test our method on videos. ACKNOWLEDGEMENT I take it a deemed privilege to express my sincere thanks to all concerned who have contributed either directly or indirectly for the successful completion of my work. It gives me immense pleasure to express my deepest sense of gratitude and sincere thanks to my respected guide Mr. Raj Kumar Sahu Associate Professor, and Mr. Abhishek Mishal, Assistant Professor, Department of Electronics and Telecommunication Engineering, Chhatrapati Shivaji Institute of Technology, Durg, Chhattisgarh, for his revered guidance throughout my dissertation work, which made this task a pleasant job. It was real pleasure to work under his guidance. REFERENCES [1]. Codruta Orniana Ancuti and Cosmin Ancuti single image dehazing by multi- scale fusion ieee transactions on image processing, vol. 22, no. 8, august [2]. Paresh Rawat, Sapna Gangrade, Pankaj Vyas Implementation of Hybrid Image Fusion Technique Using Wavelet Based Fusion Rules (IJCTEE)Volume 1, Issue 1july 12, Using Polarization Proc. Computer Vision & Pattern Recognition Vol. 1, pp (2001). [5]. Pengli LU, Qiang ZHANG Single Image Dehazing Method, Journal of Computational Information Systems 10: 4 (2014) [6]. Peter Carr, Richard Hartley Improved Single Image Dehazing using Geometry Australian National University and NICTAC anberra, Australia. [7]. A. Ben Hamzaa, Yun Heb, Hamid Krimc,and Alan Willskyd A approach to pixel-level image Fusion Integrated Computer-Aided Engineering 12 (2005) [8]. Yufeng Zheng, Multi-scale Fusion Algorithm Comparisons :Pyramid, DWT and Iterative DWT 12th International Conference on Information Fusion Seattle, WA, USA, July 6-9, [9]. N. INDHUMADHI, G. PADMAVATHI Enhanced Image Fusion Algorithm Using Laplacian Pyramid and Spatial frequency Based Wavelet Algorithm International Journal of Soft Computing and Engineering (IJSCE) ISSN: , Volume-1, Issue-5, November [10]. Yufeng Zheng, Edward A. Essock and Bruce C. Hansen An Advanced Image Fusion Algorithm Based on Wavelet Transform Incorporation with PCA and Morphological Processing. [11]. R. T. Tan, Visibility in bad weather from a single image, in Proc.IEEE Conf. Comput. Vis. Pattern Recognit, Jun. 2008, pp [12]. R. Fattal, Single image dehazing, ACM Trans. Graph., SIGGRAPH, vol. 27, no. 3, p. 72, [13]. K. He, J. Sun, and X. Tang, Single image haze removal using dark channel prior, in Proc. IEEE Conf. Comput. Vis. Pattern Recognit.,Jun. 2009, pp [14]. J. Tarel, N. Hauti, Fast visibility restoration from a single color or gray level image, Proceedings of IEEE Internation Conference on Computer Vision (ICCV), Kyoto, Japan: IEEE Computer Society, 2009, pp [3]. Jin-Hwan Kim, Jae-Young Sim, and Chang-Su Kim single image dehazing based on contrast enhancement ieee [4]. Yoav Y. Schechner, Srinivasa G. Narasimhan and Shree K. Nayar Instant Dehazing of Images [15]. Amina Saleem, AzeddineBeghdadi and Boualem Boashash, Image fusion-based contrast enhancement Saleem et al. EURASIP Journal on Image and Video Processing [16]. M.A. Mohamed1 and B.M. El-Den Implementation of Image Fusion Techniques 32

5 Using FPGA IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.5, May [17]. Dr. H.B. Kekre et al. review on image fusion techniques and performance evaluation parameters International Journal of Engineering Science and Technology (IJEST) Vol. 5 No.04 April [18]. Zheng Liu Objective Assessment Of Multiresolution Image Fusion Algorithms For Context enhancement in Night VisioN:A Comparative Study, IEEE transaction,vol 34, no.1 january2012. [19]. Shutao Li, Bin Yang Multifocus image fusion using region segmentation and spatial frequency accepted 31 October Image and Vision Computing 26 (2008) [20]. S.M.Mukane1,Y.S.Ghodake2 and P.S. Khandagle Image enhancement using fusion by wavelet transform and Laplacian Pyramid IJCSI International Journal of Computer Science Issues, Vol. 10, Issue 4, No 2, July [21]. E. H. Adelson, C. H. Anderson, J. R. Bergen Pyramid methods in image processing RCA Engineer 29-6 Nov/Dec [22]. J.P. Oakley and B.L. Satherley, Improving Image Quality in Poor Visibility Conditions Using a Physical Model for Degradation, [23]. Museum of Science, Leonardo s Perspective. mos. org/sln/leonardo/investigatingaerialp.html, [24]. M.J. Rakovic, G.W. Kattawar, M. Mehrubeoglu, B.D. Cameron, L.V. Wang, S. Rastegar, and G.L. Cote, Light Backscattering Polarization Patterns from Turbid Media: Theory and Experiment, Applied Optics, vol. 38, [25]. M.P. Rowe, E.N. Pugh Jr., J.S. Tyo, and N. Engheta, Polarization- Difference Imaging: A Biologically Inspired Technique for through Scattering Media, Optical Letters, vol. 20,