Enhancing retrieval of images on the web through effective use of associated text and semantics from low-level image features.

Content-based indexing and retrieval has emerged as an important area in computer vision and multimedia computing. Current solutions for searching image data primarily deal with associated text and low-level image features. Humans tend to use high-level concepts in everyday life; user queries are typically based on higher-level semantics and not low-level image features. However, what current computer vision techniques can automatically extract from images are mostly low-level visual features. To narrow down this semantic gap, some off-line and on-line processing is needed. The state-of-the-art image retrieval approach is to incorporate image semantics with low-level visual primitives to enhance the retrieval performance. Unfortunately the current mainstream of the image retrieval technologies in most web search engines is keyword-based retrieval; they have not explored the full potential of semantics of an image through effective use of its nearby text. Therefore I propose an image retrieval system that captures semantics of an image through effective use of its associated text and use integrated system architecture for keyword-based retrieval with low-level image features to enhance retrieval of images on the web. I have developed a new image retrieval system that enhances retrieval of images on the web through optimum. I conducted a preliminary study on collection of images obtained from HTML documents on the web. Based on my findings on text associated with the image, I have identified the textual contents of page title, image title, image alternate text, image caption and Meta tags are well related to an embedded image. These keywords lists have different significance in identifying the image semantics. I comparatively evaluate the performance of each keyword list exclusively to study their impact on overall retrieval effectiveness. The major contribution of my work included a full-scale development and implementation of the new image retrieval system I-Search. The system was based on an enhanced image representation that exploits the vast power of image semantics from the text associated with the images and higher-level semantic categories based on low-level image features of the images. The user-interface was designed to allow the user to communicate keywords based query and semantic categories to the image retrieval system. The performance of this new image retrieval system I-Search was compared with GoogleTM and YahooTM. Our analysis of this experiment confirmed that the integration of text associated with an image and low-level image features will lead to efficient retrieval system for content-based indexing of images on the web and will in fact substantially enhance the image searching capabilities on the web