An optimized mksift and cluster based Cross indexing approach for effective Image search and retrieval

Abstract

Due to the increasing demand for improved image indexing and etrieval approaches, image retrieval has become one of the important subjects in areas, like object recognition and artificial intelligence. Especially, Image retrieval on large scale image databases has attained more attentions, where apping features into binary codes is showing great advancement. Most of the approaches represent the image using invariant local features adopting Scale Invariant Feature Transform (SIFT). In this research, two techniques such as, MKSIFT+ Cross indexing and MKSIFT+ cluster indexing have been developed for retrieving the suitable images from the database based on the query. In MKSIFT+ Cross indexing, a new approach for image retrieval, Multiple Kernel SIFT (MKSIFT) that extracts the features from the pre-processed input image. It utilizes the following steps of SIFT: detection of extrema, key point removal, orientation assignment, and calculation of descriptor, to extract the feature points. MKSIFT computes the keypoint descriptor with the introduction of exponential and tangential kernels, where the weights assigned to the kernels are selected by Particle Swarm-Fractional Bacterial foraging optimization (PS-FBFO) algorithm. Moreover, it performs a cross-indexed image search by converting the feature points of MKSIFT into binary codes. The pre-processing is carried out to enhance the quality of the image by making the feature extraction process more reliable. Pre-processing helps the image database to attain more relevant images. It is required that, all the images that are to be processed should have equivalent height and width, even though the image can be of any size based on the acquisition device settings. The location and the scales of the images are identified by determining the stable features over all scales using the scaling function called scale space. Accordingly, the scale space is computed using the convolution of a Gaussian in the input image. Scale-space extrema is utilized in difference-of-Gaussian (DoG) function to detect the keypoint localization by convolving the function with the image. newline