Implementation and Evaluation of Query by Sketch (Master Thesis, Finished)
With the rapid growth of digital image data available, the need for efficient image retrieval techniques has tremendously increased. As annotation-based image retrieval can be very time consuming and unreliable, research in content-based image retrieval (CBIR) has become more and more important and widespread. Within CBIR Query by Sketch (QbS) is a small field of research but can be expected to provide valuable solutions with the growth of more sophisticated input technologies like digital pens and paper.
This Master Thesis describes the implementation and extension of an existing approach to QbS: the Angular Radial Partitioning (ARP). This approach compares the amount of edge pixels in the partitioned black and white sketch and the target images. However ARP only provides support for a full image similarity search and is not able to find images from which only a partial sketch is available. We propose the application of bounding boxes, image regions and an adapted distance measure, to overcome this weakness. With the usage of the Average Normalized Modified Retrieval Rank (ANMRR) this improvements are evaluated.
Start / End Dates
2009/03/15 - 2009/09/14