Minakshi Banerjee

Content-based image retrieval using visually significant point features

This paper presents a new image retrieval scheme using visually significant point features. The clusters of points around significant curvature regions (high, medium, and weak type) are extracted using a fuzzy set theoretic approach. Some invariant color features are computed from these points to evaluate the similarity between images. A set of relevant and non-redundant features is selected using the mutual information based minimum redundancy-maximum relevance framework. The relative importance of each feature is evaluated using a fuzzy entropy based measure, which is computed from the sets of retrieved images marked relevant and irrelevant by the users. The performance of the system is evaluated using different sets of examples from a general purpose image database. The robustness of the system is also shown when the images undergo different transformations.

Edge based features for content based image retrieval

The common problem in content based image retrieval (CBIR) is selection of features. Image characterization with lesser number of features involving lower computational cost is always desirable. Edge is a strong feature for characterizing an image. This paper presents a robust technique for extracting edge map of an image which is followed by computation of global feature (like fuzzy compactness) using gray level as well as shape information of the edge map. Unlike other existing techniques it does not require pre segmentation for the computation of features. This algorithm is also computationally attractive as it computes different features with limited number of selected pixels.

Content based image retrieval with fuzzy geometrical features

This paper presents a robust technique for Content Based Image Retrieval (CBIR) using fuzzy edge map of an image. Fuzzy compactness vector is computed from fuzzy edge map thresholded at different levels of the unsegmented image, which also incorporates gray level contrast information embedded in the edges. The resemblance of two images is defined as the similarity between the computed feature vectors.

Handling of impreciseness in gray level corner detection using fuzzy set theoretic approach

Reliable corner detection is an important task in determining shape of different regions in an image. To detect corners in a gray level image under imprecise information, an algorithm based on fuzzy set theoretic model is proposed. The uncertainties arising due to various types of imaging defects such as blurring, illumination change, noise, etc., usually result in missing of significant curvature junctions (corners). Fuzzy set theory based modeling is well known for efficient handling of impreciseness. In order to handle the incompleteness arising due to imperfection of data, it is reasonable to model image properties in fuzzy frame work for reliable decision making. The robustness of the proposed algorithm is compared with well known conventional detectors. The performance is tested on a number of benchmark test images to illustrate the efficiency of the algorithm.

Corner detection using support vector machines

A support vector machine based algorithm for corner detection is presented. It is based on computing the direction of maximum gray-level change for each edge pixel in an image, and then representing the edge pixel by a four dimensional feature vector constituted by the count of other edge pixels lying in a window centred about and having each of the possible four directions as their direction of maximum local gray-level change. A support vector machine is designed using this feature vectors and the support vectors, representing critical points in a classification problem, correspond to the corner points. The algorithm is straightforward and does not involve computation of complex differential geometric operators. It has implicit learning capability resulting in good performance for a wide range of images.

Interactive image retrieval with wavelet features

This paper presents an iterative Content Based Image Retrival( CBIR) system with Relevance Feedback (RF), in which M-band wavelet features are used as representation of images. The pixels are clustered using Fuzzy C-Means (FCM) clustering algorithm to obtain an image signature and Earth Movers Distance (EMD) is used as a distance measure. Fuzzy entropy based feature evaluation mechanism is used for automatic computation of revised feature importance and similarity distance at the end of each iteration. The performance of the algorithm is tested on standard large multi-class image databases and compared with MPEG-7 visual features.

Retinal vessel segmentation using Gabor filter and artificial neural network

This paper demonstrates an automated segmentation scheme of retinal vasculature using Gabor filter bank, which is optimized on the basis of entropy. Different distributions of filter responses are encoded into features and the vasculature of normal and abnormal retina are segmented by artificial neural network(ANN). The training set of labeled pixels is obtained from the ground truth images of DRIVE database. The Receiver Operating Characteristics (ROC) in both abnormal and normal cases shows 96.16% accuracy.

Image retrieval using fuzzy relevance feedback and validation with MPEG-7 content descriptors

Content-Based Image retrieval has emerged as one of the most active research directions in the past few years. In CBIR, selection of desired images from a collection is made by measuring similarities between the extracted features. It is hard to determine the suitable weighting factors of various features for optimal retrieval when multiple features are used. In this paper, we propose a relevance feedback frame work, which evaluates the features, from fuzzy entropy based feature evaluation index (FEI) for optimal retrieval by considering both the relevant as well as irrelevant set of the retrieved images marked by the users. The results obtained using our algorithm have been compared with the agreed upon standards for visual content descriptors of MPEG-7 core experiments.

Tackling Curse of Dimensionality for Efficient Content Based Image Retrieval

This paper proposes a content based image retrieval (CBIR) technique for tackling curse of dimensionality arising from high dimensional feature representation of database images and search space reduction by clustering. Kernel principal component analysis (KPCA) is taken on MPEG-7 Color Structure Descriptor (CSD) (64-bins) to get low-dimensional nonlinear-subspace. The reduced feature space is clustered using Partitioning Around Medoids (PAM) algorithm with number of clusters chosen from optimum average silhouette width. The clusters are refined to remove possible outliers to enhance retrieval accuracy. The training samples for a query are marked manually and fed to One-Class Support Vector Machine (OCSVM) to search the refined cluster containing the query image. Images are ranked and retrieved from the positively labeled outcome of the belonging cluster. The effectiveness of the proposed method is supported with comparative results obtained from (i) MPEG-7 CSD features directly (ii) other dimensionality reduction techniques.

Content-Based image retrieval using wavelet packets and fuzzy spatial relations

This paper proposes a region based approach for image retrieval. We develop an algorithm to segment an image into fuzzy regions based on coefficients of multiscale wavelet packet transform. The wavelet based features are clustered using fuzzy C-means algorithm. The final cluster centroids which are the representative points, signify the color and texture properties of the preassigned number of classes. Fuzzy Topological relationships are computed from the final fuzzy partition matrix. The color and texture properties as indicated by centroids and spatial relations between the segmented regions are used together to provide overall characterization of an image. The closeness between two images are estimated from these properties. The performance of tlie system is demonstrated using different set of examples from general purpose image database to prove that, our algorithm can be used to generate meaningful descriptions about the contents of the images.