Sketch Based Image Retrieval Using Watershed Transformation

Abstract

In the field of Digital Image Processing Content Based Image Retrieval is becoming very popular. Google and Yahoo have tools on Digital Image Processing. They are known to be Google Images and Yahoo! Images Search. They are based on textual annotation of images. In textual annotations with the help of keywords images are retrieved. This is not very much effective approach as their performances are not satisfactory. The content based Image retrieval is based on automatic extraction of content based on color, texture, etc. The focus of this report is to tell about the obstacles in development of Content Based Image Retrieval which is based on free hand sketch(Sketch Based Image Retrieval). The focus will be to try and create task specific descriptor to handle informational gap that exists between coloured images and sketches which will give opportunity for effective search. The descriptor is constructed after such special sequence of pre-processing steps that the sketch and transformed images can be compared. EHD, HOG and SIFT are the topics that we have covered. Overall, Results have shown that sketch based system allows intuitive access to search tools. Digital Libraries, Crime prevention, Photo sharing websites, etc are some of the targets applications where SBIR can work. Identifying victims in forensics and law enforcements or apprehending suspects can be great value for this system. Matching a sketch of cup shot can be a good example in forensics. Sketching or visual content of querying a picture and then returning an image is intensified recently, In the area of image processing it will require a wide methodology spectrum.