IOSR Journal of Electronics and Communication Engineering (IOSR-JECE) e-ISSN: 2278-2834,p- ISSN: 2278-8735.Volume 10, Issue 4, Ver. II (Jul - Aug .2015), PP 72-75 www.iosrjournals.org DOI: 10.9790/2834-10427275 www.iosrjournals.org 72 | Page Shadow Detection and its Removal from Images Using Strong Edge Detection Method Prateek Sharma 1 , Reecha Sharma 2 1 (M.Tech Student, Deptt. Of Electronics and Communication Engineering, Punjabi University, Patiala, India) 2 (Assistant Prof., Deptt. Of Electronics and Communication Engineering, Punjabi University, Patiala, India) Abstract: Shadows cause problems in image processing. In this paper, a new methodology for shadow removal based on Strong Edge Detection (SED) method is proposed. The strong shadow edges are recognized by learning the patch based characteristics of shadow edges and then image features are analyzed to guide a Shadow Edge Classifier. Also, spatial patch smoothing is used to enforce uniformity between adjacent patches. The entropy and standard deviation results of both earlier Patch based Shadow Edge Detection Method and proposed SED method are calculated and compared with each other. The results show that the proposed SED method is better than the previous Patch Based Shadow Edge Detection method. Keywords: Strong Edge Detection, Shadow Removal, Shadow Edge Classifier, Shadow Edges, Shadow Removal I. Introduction Eliminating Shadows from images is a complicated task. To get a high-quality shadow-free image that is a replica of a true shadow-free scene is even more difficult. Different lighting circumstances, variety of shadowed surfaces etc affects shadows in images. Furthermore, shadow regions may undergo contrast enrichment, which may introduce perceptible artifacts in the images without shadows. The shadowing effect is compounded in region where there are striking changes in surface elevation mostly in urban areas [1]. The obstruction of light by objects creates shadows in a scene. The shadow areas are less illuminated than the surrounding areas. The shadows are classified as hard and soft on the basis of intensity. The soft shadows retain the texture of the background surface, whereas the hard shadows are too dark and have little texture [2]. In Satellite images, shadows occur due to imaging conditions and the presence of various high-rise objects and this is mostly in urban areas [3]. In moving shadows the real challenge is to classify moving shadow points which are many times misclassified as moving object points in a video sequences causing problems in vision applications. Moving cast shadows deform the evaluation of shape and color characteristics of target objects [4, 11]. For creating virtually realistic worlds and visualization applications, Front-projection displays are being used. However, these systems have drawback that Users and different objects present in the environment can easily and involuntarily obstruct projectors, creating shadows on the displayed image [5]. Though in some cases, the shadows provide useful information, such as the relative position of an object from the source but they create complexities in visualization applications like segmentation, detecting objects and object classification [6]. An object sometimes cast a shadow i.e. self-shadow on itself. Self-shadows in comparison to hard shadows have more brightness. Usually Cast shadow comes under hard shadows, divided into umbra and penumbra region [7], 8]. The detection of hard shadows is complicated as they may be flawed as dark objects rather than shadows [9]. Also it is difficult to distinguish dark objects and shadows from a single image. As shadows are time dependant and possess seasonal characteristics, variation of shadow casting condition alters the image [12]. Thus to recognize shadows and to eliminate them always remain a pre-requisite task. This paper explains the work carried for detecting and removing shadows. This paper has been organized into five sections, described as: Section I defining the basic introduction, Section II includes basics of feature extraction technique. Feature extraction technique is followed by Section III containing proposed methodology. This section explains different steps that are executed to get the end results. In sections IV & V, results and conclusions are drawn respectively. II. Basics Of Feature Extraction When the input to an algorithm is very large to be analyzed and it is redundant (e.g. repetitiveness of images presented as pixels), then it is transformed to limited features (portions or shapes of an image). This phenomenon is feature Extraction. The extracted features will contain significant information from the input and task is carried out.