Nishat Ahmad

An intrinsic semantic framework for recognizing image objects

The paper proposes a new approach to find semantic meanings in visual object class structure, in line with the Gestalt laws of proximity. Micro level semantic structures are formed by line segments (arcs also approximated into line segments based on pixel deviation threshold) which are in close proximity. These structures are hierarchically combined till a semantic label can be assigned. The algorithm extracts semantic groups, their inter-relations and represents these using a graph. Invariant geometrical properties of the groups and relations are used as vertex and edge labels. A graph model captures the inter class variability by analyzing the repetitiveness of structures and relations and uses it as a weighting factor for classification. The algorithm has been tested on a standard benchmark database and compared with existing approaches.

Defining a Set of Features Using Histogram Analysis for Content Based Image Retrieval

A new set of features are proposed for Content Based Image Retrieval (CBIR) in this paper. The selection of the features is based on histogram analysis. Standard histograms, because of their efficiency and insensitivity to small changes, are widely used for content based image retrieval. But the main disadvantage of histograms is that many images of different appearances can have similar histograms because histograms provide coarse characterization of an image. Hence we further refine the histogram using the histogram refinement method. We split the pixels in a given bucket into several classes just like histogram refinement method. The classes are all related to colors and are based on color coherence vectors. After the calculation of clusters using histogram refinement method, inherent features of each of the cluster is calculated. These inherent features include size, mean, variance, major axis length, minor axis length and angle between x-axis and major axis of ellipse for various clusters.

Histogram Based Corner Angle Representation for Object Retrieval

The paper presents an algorithm which uses corner and intersecting lines angle information with the corner boundary pixel information to construct a feature vector for representing an image and subsequently for searching the similar images from a database. The corners are extracted by finding the intersections of the detected lines found using Hough transform. Taking the corner as the center coordinate, the angles of the intersecting lines are determined and inner and outer object boundary pixel information around the corner is used to represent the corner geometry. For similarity measurement, the histogram of the information is used. This result in a significant small size feature matrix compared to the algorithms using color features. Experimental results show that it is computationally efficient in similarity measurement and the image corners being noise invariant produce good results in noisy environments.

Object Retrieval Approach with Invariant Features Based on Corner Shapes

This paper presents a new technique for corner shape based object retrieval from a database. The proposed feature matrix consists of values obtained through a neighborhood operation of detected corners. This result in a significant small size feature matrix compared to the algorithms using color features and thus is computationally very efficient. The corners have been extracted by finding the intersections of the detected lines found using Hough transform. As the affine transformations preserve the co-linearity of points on a line and their intersection properties, the resulting corner features for image retrieval are robust to affine transformations. Furthermore, the corner features are invariant to noise. It is considered that the proposed algorithm will produce good results in combination with other algorithms in a way of incremental verification for similarity.

Image Feature Vector Construction Using Interest Point Based Regions

The paper presents a new approach for content based retrieval of images. The algorithm uses information sampled from around detected corner points in the image. A corner detection approach based on line intersections has been employed using Hough transform for line detection and then finding intersecting, near intersecting or complex shaped corners. As the affine transformations preserve the co-linearity of points on a line and their intersection properties, the corner points obtained as such retain the much desired property of repeatability and hence ensure the similar pixel samples under various transformations and are robust to noise. K-means clustering algorithm is used to assign class labels to the extracted sample mean and variance of the corner regions from a random selection of training images and used for learning a Gaussian Byes classifier to classify whole training image database. Histogram of the class members in an image has been used as a feature vector. The retrieval performance and behavior of the algorithm has been tested using four different similarity measures.

Content Based Image Retrieval Using Localized Line Segment Groupings

The paper presents a new approach for feature representation using semantic line groupings in an image. The algorithm uses the hypothesis in line with Gestalt laws of proximity that as a baseline in an image, semantic structures are formed by line segments placed in close proximity to each other. The algorithm uses line segments in an image to form semantic groups based on a minimum distance threshold. The semantic line groupings are differentiated from each other by the number of group members and their geometrical properties represented as histograms. The results are analyzed using different similarity measures to understand the strengths and weaknesses of the grouping approach and those of the similarity measures.

Corner geometry representation using code vectors for image retrieval

The paper presents an algorithm which uses code vectors to represent corner geometry information for searching the similar images from a database. The corners have been extracted by finding the intersections of the detected lines found using Hough transform. Taking the corner as the center coordinate, the angles of the intersecting lines are determined and are represented using code vectors. A code book has been used to code the each corner geometry and indexes to the code book are generated. For similarity measurement, the histogram of the code book indexes is used. This result in a significant small size feature matrix compared to the algorithms using color features. Experimental results show that use of code vectors is computationally efficient in similarity measurement and the corners being noise invariant produce good results in noisy environments.

Detecting collisions in an unstructured environment through path anticipation

Detecting and avoiding a potential collision is the most important aspect when we talk about mobile objects of all types, ranging from autonomous vehicles, ship and aircraft navigation to robot manipulators etc. Image processing techniques could be used to provide solutions in such scenarios. Detection of a possible collision requires accurate information about the path of a moving object. The projected paths of all mobile objects in vicinity can illustrate if a possible collision scenario exists and its level of seriousness. Based on which, collision avoidance actions can be initiated. This paper presents use of image processing techniques for collision detection and avoidance scenarios using the statistical measurements