The Research Of Image Retrieval Based On Sketch

With the rapid development of internet and multimedia technology, there has been an explosive increase in image information,which make massive image database to become the main target for people to access and retrieve images.How to retrieve the expected information quickly and effectively among the massive image database becomes one of the most hot research field of Image retrieval.The Research of Image Retrieval based on Sketch is an important research field of CBIR.During the early period,image retrieval is mainly based on text. With touch screen technology applicated in smartphones and tablet computers is becoming more and more.. Part of the traditional mechanical of the keyboard function replaced by smoothly touch screen experience gradually,The human-computer interaction has become the main input mode.People can draw the sketch of the things in their mind on the touch screen.As a kind of expression way of human thought and vision,hand-drawn sketches can not only express the semantic demand of the user, but intuitived.The technology has attracted much attention.This retrieval methods which is mainly to calculate the feature similarity between images by using low-level features that are extracted from images,and finally get the images which are very similar.This approach is simple and straightforward,But the available information only shape information. In order to achieve a higher retrieval efficiency, it is necessary to extract effective bottom shape characteristics, and then use the appropriate model to express these characteristics.This thesis focuses on the automatic semantic image annotation technologies.The main work and innovations are as follows:(1) An effective sketch retrieved method is proposed based on Combining with The information entropy and The image potential energy.First,we divided the area which surrounding block of target pixels;Then obtain its potential energy according to the characters of each pixel location;Finally we combined with the feature of entropy and the potential energy to achieve sketch retrieval.The experimental results show that the method have good performance on the recognition rate in large scale image set, to verify the effectiveness.(2)An effective sketch retrieved method is proposed based on Combining with Histogram of Oriented Gradient and The information entropy.First,we divided the area which surrounding block of target pixels;Then obtain its Histogram of Oriented Gradient according to the characters of each pixel location;Finally we combined with the feature of entropy and the potential energy to achieve sketch retrieval.The experimental results show that the method have good performance on the recognition rate in a certain size image set.