A Sketch based Image Retrieval with Descriptor based on Constraints

To match sketch and real images some processing is required, because it is very difficult to directly match sketch and real image. Real images contain noise due to many reasons, which makes it very difficult for matching. For this descriptor is designed so that it can give best match by finding relationships between edges and line segments. By applying edge length as constraint, the retrieval performance is increases. Proposed framework tested on public datasets and results shows that proposed method improves SBIR performance significantly.