M Tech Dissertations
Permanent URI for this collectionhttp://ir.daiict.ac.in/handle/123456789/3
Browse
6 results
Search Results
Item Open Access Image ranking based on clustering(Dhirubhai Ambani Institute of Information and Communication Technology, 2011) Sharma, Monika; Mitra, Suman K.In a typical content-based image retrieval (CBIR) system, query results are a set of images sorted by feature similarities with respect to the query. However, images with high feature similarities to the query may be very different from the query. We introduced a novel scheme to rank images, cluster based image ranking, which tackle this difference in query image and retrieved images based on hypothesis: semantically similar images tends to clustered in same cluster. Clustering approach attempts to capture the difference in query and retrieved images by learning the way that similar images belongs to same cluster. For clustering color moments based clustering approach is used. The moment is the weighted average intensity of pixels. The proposed method is to compute color Moments of separated R,G,B components of images as a feature to get information of the image. This information can be used further in its detail analysis or decision making systems by classification techniques. The moments define a relationship of that pixel with its neighbors. The set of moments computed will be feature vector of that image. After obtaining the feature vector of images, k-means classification technique is used to classify these vectors in k number of classes. Initial assignment of data to the cluster is not random, it is based on maximum connected components of images. The two types of features are used to cluster the images namely: block median based clustering and color moment based clustering. Experiments are performed using these features to analyze their effect on results. To demonstrate the effectiveness of the proposed method, a test database from retrieval result of LIRE search engine is used and result of Lire is used as base line. The results conclude that the proposed methods probably give better result than Lire result. All the experiments have been performed on in MATLAB(R). Wang database of 10000 images is used for retrieval. It can be downloaded from http://wang.ist.psu.edu/iwang/test1.tarItem Open Access Practical approach for depth estimation and image restoration using defocus cue(Dhirubhai Ambani Institute of Information and Communication Technology, 2011) Ranipa, Keyur R.; Joshi, Manjunath V.Reconstruction of depth from 2D images is an important research issue in computer vision. Depth from defocus (DFD) technique uses space varying blurring of an image as a cue in reconstructing the 3D structure of a scene. In this thesis we explore the regularization based approach for simultaneous estimation of depth and image restoration from defocused observations. We are given two defocused observations of a scene that are captured with di erent camera parameters. Our method consists of two steps. First we obtain the initial estimates for the depth as well as for the focused image. In the second step we re ne the solution by using a fast optimization technique. Here we use the classic depth recovery method due to Subbarao for ob- taining the initial depth map and Wiener lter approach for initial image restoration. Since the problem we are solving is ill-posed and does not yield unique solution, it is necessary to regularize the solution by imposing additional constraint to restrict the solution space. The regularization is performed by imposing smoothness constraint only. However, for preserving the depth and image intensity discontinuities, they are indenti ed prior to the minimization process from initial estimates of the depth map and the restored image. The nal solution is obtained by using computationally e client gradient descent algorithm, thus avoiding the need for computationally taxing algorithms. The depth as well as intensity edge details of the nal solution correspond to those obtained using the initial estimates. The experimental results indicate that the quality of the restored image is found to be satisfactory even under severe space-varying blur conditions.Item Open Access Collusion resistant fingerprinting(Dhirubhai Ambani Institute of Information and Communication Technology, 2011) Juneja, Sandeep; Raval, Mehul S.Digital watermarking is used to carry information by embedding information into the cover data in a perceptually visible or non visible manner. In today's sea of digital information, there are many problems associated like identi cation of the owner of content, and detection of authorized content receivers. Digital ngerprinting, one of the application of watermarking, is one way to detect authorized content receivers from illegally redistributed media. One powerful scheme to broke digital ngerprint scenario is `collusion attack' in which users share information to remove their embed- ded ngerprint. In this research work, we have proposed a ngerprint technique that is robust against average collusion attack and has capability to trace colluders for images. Independent ngerprints are randomly generated using independent and identically distributed (IID) Gaussian source. We proposed two schemes. In rst scheme, n- gerprints were embedded using additive embedding rule and spread spectrum (SS) technique. This scheme is based on embedding ngerprint in di erent block of discrete cosine transformation (DCT). In second, ngerprints were embedded in independent components (ICs) generated by applying independent component analysis (ICA) on cover image. In both schemes, we used non-blind watermarking and correlation based detector. The result shows that the schemes are robust against average collusion at- tack.Item Open Access Object segmentation in still camera videos(Dhirubhai Ambani Institute of Information and Communication Technology, 2010) Pandya, Sweta A.; Mitra, Suman K.The goal of object segmentation is to simplify and change the representation of an image into more meaningful so that it can easily analyse. Segmentation is the process of partitioning the digital image into multiple segments (set of pixels). It is the foremost step before performing other operations like recognition, scene understanding, tracking, etc. Main purpose of video segmentation is to extract the objects of interest from a series of consecutive video frames. For example surveillance video requires high-level image understanding and scene interpretation for tracking the special events. Another example is of segmenting flower from an image and video in which there are variety of flowers, the variability within a particular flower, and the variability of the imaging conditions – lighting, pose, etc. There are various approaches for segmenting the object from an image. Some of them are histogram based approach, region based approach and graph partitioning approach. In graph partitioning approach, the image being segmented is modelled as a weighted, undirected graph. Each pixel is represented as a node in the graph, and an edge is formed between every pair of pixels. The weight of an edge is a measure of the similarity between the pixels. Some popular algorithms of graph partitioning category are random walker, minimum mean cut, minimum spanning tree-based algorithm and normalized cut. In graph partitioning approach, the normalized cut algorithm is used to solve the grouping problem. In this algorithm, image is partitioned into disjoint sets by removing the edges connecting the segments. The partition can be done by finding the splitting point. The optimal solution of the splitting point is computed by solving the Eigen value problem. The optimal partitioning of the graph is the one that minimizes the weights of the edges that were removed. A normalized cut criterion measures the dissimilarity between the different groups as well as total similarity within the groups. Here group size doesn’t matter for normalized cut criterion. The normalized cut can be computed using three different splitting points and the result is analysed accordingly. A common approach for detecting the object from a video is to perform background subtraction, which identifies moving objects from the portion of a video frame that differs significantly from a background model and then apply the segmentation algorithm to that video. Here background subtraction has been done by frame difference method. In this method previous frame is subtracted from the current frame and difference is compared with the specific threshold value. For experimental purpose, videos of different flowers and movement of the tennis balls have been taken. All the experiments have been performed on both gray scale image and videos in MATLAB.Item Open Access New learning based super resolution using contourlet transform(Dhirubhai Ambani Institute of Information and Communication Technology, 2009) Singh, Vineet P.; Joshi, Manjunath V.new learning based super-resolution reconstruction using contourlet transforms is proposed. contourlet transform provides high degree of directionality. It captures geometrical smoothness along multiple directions and learns the edges present in an image normal to the contour. For learning purpose, training set of low resolution (LR) and high resolution (HR) images, all captured using the same camera, are used. Here two and three level contourlet decomposition for LR images (test image and training image dataset) and HR training images respectively. The comparison of contourlet coeffcients of LR test image from the LR training set using minimum absolute difference (MAD) criterion to obtain the best match contourlet coeffcient. The finer details of test image are learned from the high resolution contourlet coefficients of the training data set. The inverse contourlet transform gives super resolved image corresponding to the test image.Item Open Access Content-based image retrieval system for multi-object images using a combination of features(Dhirubhai Ambani Institute of Information and Communication Technology, 2006) Katare, Aradhana; Mitra, Suman K.; Banerjee, AsimContent-based image retrieval (CBIR) is a research area dedicated to address the retrieval of images based on automatically derived features from the content of the images in database. Traditional CBIR systems generally compute global features of the image for example, based on color histograms. When a query images is fired, it returns all those images whose features match closely with the query image. The major disadvantage of such systems based on global features is that they return the images that match globally but cannot possibly return images corresponding to some particular objects in the query image. The thesis addresses this problem and proposes a CBIR system for multi object image database with 3D objects using the properties of the object in the images for retrieval. Object segmentation has been achieved using GVF Active Contour. An inherent problem with active contours is initialization of contour points. The thesis proposes an approach for automatic initialization of contour points. Experimental results show that the proposed approach works efficiently for contour initialization. In the thesis in addition to shape feature using modified chain code other features for object retrieval using colour with the aid of colour moments and texture using Gabor Wavelets have also been used. A comparative study has been made as to which combination of features performs better. Experimental results indicate that the combination of shape and color feature is a strong feature for image retrieval.