An Image Indexing and Region based on Color and Texture

Abstract

From the previous decade, the enormous rise of the internet has tremendously maximized the amount image databases obtainable. This image gathering such as art works, satellite and medicine is fascinating ever more customers in numerous application domains. The work on image retrieval primarily focuses on efficient and effective relevant images from huge and varied image gatherings which is further becoming more fascinating and exciting. In this paper, the author suggested an effective approach for approximating large-scale retrieval of images through indexing. This approach primarily depends on the visual content of the image segment where the segments are obtained through fuzzy segmentation and are demonstrated through high-frequency sub-band wavelets. Furthermore, owing to the complexity in monitoring large scale information and exponential growth of the processing time, approximate nearest neighbor algorithm is employed to enhance the retrieval speed. Thus, a locality-sensitive hashing using (K-NN Algorithm) is adopted for region-aided indexing technique. Particularly, as the performance of K-NN Approach hinges essentially on the hash function segregating the space, a novel function was uncovered motivated using E8 lattice which could efficiently be amalgamated with multiple probes K-NN Approach and query-adaptive K- NN Approach. To validate the adopted hypothetical selections and to enlighten the efficiency of the suggested approach, a group of experimental results associated to the region-based image retrieval is carried out on the COREL data samples.