The Research And Implementation Of Sketch-based Image Retrieval

With the popularity of the Internet and the rapid development of multimedia technology, the number of images in the the Internet has increased dramatically, the scholars have also been attracted by image retrieval technology, and achieved remarkable results. In recent years, with the popularity of smart mobile phone, notebook computer and other related equipment, the rapid development of touch screen technology, the technology of sketch-based retrieval has gradually attracted the domestic and foreign scholars.Device has a touch screen functions as a new human-computer interaction, making human-computer interaction becomes increasingly smooth. By devices with a touch function, different people can easily be in the minds of objects and simple lines sketched out in the image library retrieved similarly shaped objects. However, due to variability in the line of hand-drawn sketches and paintings of different people based on uncertainty and subjective factors that caused extremely difficult feature representation, while making a low recognition rate of hand-drawn sketches. Sketch features generally implicit in the shape of the object, as there is a big difference between different users, so the effect of hand-drawn sketches of retrieval is largely characterized by the quality of the child’s description of the decision, a feature descriptor can be reduced if differences between different users will get a good description of the object recognition rate, which is also in the process of hand-drawn sketches retrieval difficult.This paper describes the background and significance of retrieval based on hand-drawn sketches, and then summarizes the current research freehand sketches retrieval technology. Analysis method of extracting features several hand-drawn sketches, and then put forward a skeleton and features hand-drawn sketches based on the minimum bounding quadrilateral descriptor extraction method. Firstly, by extracting the information dot of the image pre-processing, and then extracted four vertices starting skeleton, the branch point and the minimum circumscribed quadrangle as feature points, then each extracted feature point with respect to the entropy of the information point. Experiments show that the method on different image sets have achieved good results. Then we propose a hand-drawn sketch retrieval systems that use the "bag of words model" to represent sketch user input. The system extracts the characteristics of the user hand-drawn sketches, and then identity the nearest five classes to get a hand-drawn sketch of the label, and then use the five labels in existing search by tags similar color pictures on the search for the search engines.