Towards Semantic Image Search

Internet has long been a part of our life. In the recent days, image and video based search has also become very popular. In this work, we propose a novel web based image search algorithm which will overcome the shortcomings of the present popular search algorithms. While the current algorithms mainly use the text tag based information for image search, we combine image and text based information for more accurate results. Our text analysis engine investigates the associated text with a webpage image and returns the possible object classes present in the image. Finally, our image recognition engine identifies the maximal probable classes given the image features and extracted textual information.