Text based annotation and integrating visual features for CBIR system in Deep Learning

Abstract

In content-based image retrieval system, the user’s query results are a set of images sorted according to the feature similarities between the images with respect to the query. In the real world, the images are main information source and represents features of objects like their color, shape and other attributes. From the various domains like fashion, crime prevention and medicine etc, we need some efficient image searching, browsing and retrieval tools. Although lots of work has been done for indexing and retrieval of images, but still there is no universally accepted feature extraction, retrieval and indexing technique available which is both accurate and fast. In this paper we have presented an algorithm which analyzes an image based on objects and RGB components and give similar images as the output. The images will be stored in the database with a certain tag that describes it. These tags can be used for text based image retrieval and helps in visual feature extraction. Most of the current search engines use tag based image retrieval which is time consuming and less accurate in user’s perspective. We have overcome these problems in proposed algorithm. The experimental results, showing the effectiveness of algorithms are presented.