Paper Title
High Level Semantic Image Retrieval Algorithm For Corel Database

Abstract
Image retrieval (IR) is an extremely active research area since ages. It is used in variety of applications like art, photojournalism, medical and scientific applications, on-line learning, and even for personal visual information search. In large, varied collection of images, users face a problem of retrieving images relevant to the user query. An image is said to be relevant if it’s related to the query, semantically and content-wise. Bridging the semantic gap between user-provided text queries and images is challenging. Various techniques are adopted to solve this problem. Any new approach to deal with improving the relevancy rate in searches has its focus on semantic gap which exist between the low-level visual features of images and high-level textual queries. This paper proposes a scheme to reduce the semantic gap and thereby, achieve high retrieval accuracy. Semantic template is the ‘representative’ feature of a concept, calculated from a collection of sample images. Semantic template technique is employed to bridge the ‘semantic gap’. Semantic gap reduction can be achieved by maintaining a database of images and providing the user with a choice of selecting a query image along with the semantic query (keyword). For each query, feature extraction is done using GLCM method and the keyword is also transformed to a unique code. Then, the image feature vectors are combined to generate a semantic template. The same is done for the whole database and the semantic templates are compared using similarity measures. Precision of about 80% is acquired on retrieval of images using the proposed system, for the user provided query. Our approach improves the retrieval results by filtering those images that are not semantically consistent with the queries. The main idea of this paper is to point a new direction on handling the text-to-image search based on how humans identify relevant images. Keywords- Image Retrieval, CBIR, Semantic Gap, Semantic Template, GLCM.