Certain investigations on image retrieval using statistical methods and hybrid optimization algorithms

Abstract

The tremendous growth in the number of digital images has motivated the need for improvement in search and retrieval of images from a large database. This can be achieved by a technique known as Content Based Image Retrieval. Image retrieval is a technique to search for the most visually similar images to a given query image from a large image database. The major advantage of this approach is that little human intervention is required. The biggest challenge faced in this process is retrieval of the desired images from a large database with maximum precision and minimum retrieval time. Hence, an efficient image retrieval system is needed to provide a solution to the user to retrieve the required images from a large database with minimum retrieval time and high accuracy. The two important factors in image retrieval are feature extraction and similarity measurement. This thesis has addressed the challenges and problems in the areas of feature extraction and similarity measurement towards better retrieval of images from large database. Image retrieval utilizes the visual contents of an image such as color, texture, and shape in order to represent the image. The feature extraction is an important step and the effectiveness of a Content Based Image Retrieval system depends on the method of feature extraction. Selecting appropriate features play a significant role in improving the performance of image retrieval systems. In this research work, a higher order Gray Level Co-occurrence Matrix is used for extracting texture features, a new Hue-Saturation-Value model color moments are used for extracting color features, Improved Zernike Moments are used for extracting shape features and Speeded Up Robust Features are used for object detection. newline