ROUGH SKETCH-BASED IMAGE-INFORMATION RETRIEVAL

Authors propose a new visual interface called QVE (Query by Visual Example), which seeks to improve the interactive interface between human and machine for multimedia databases. The main characteristic of QVE is its ability to access the database using visual data, such as the shape of an image. A user has only to draw a rough sketch to retrieve the original image and similar images. The system evaluates the similarity between the rough sketch, i.e.. visual example, and each part of the image data in the database automatically. QVE is a very effective approach to content-based image retrieval. In this paper authors present the motivation for, and the general framework of visual interaction. Next, they propose the concept of QVE, for content-based image retrieval as a visual interaction tool. Then, they show fundamental algorithms used for QVE. Finally, they discuss some experimental results of QVE and their current evaluation of it.