Research Of Image Retrieval Technology Based On Shape Feature

As one of the most important low-level features of object, shape carries certain semantic information compared to color, texture and position relationship. Shape based image retrieval is an important part of content based image retrieval, which has wide application fields including industry design, visual communication, digital entertainment and web development, etc.Drawing is a kind of innate capacity of human. It’s one of the most common and familiar methods for people to make use of hand sketch to design or express their intentions. Nowadays, hand sketch has become one of the most natural ways of human computer interaction with the increasing popularization and application of touch-screen devices. Aimed at shape based image retrieval, this thesis mainly researches the extraction and recognition of contour feature and builds a freehand sketch retrieval system. The major work is as follows:(1) An improved shape retrieval algorithm based oncentroid distance ratio and shape context is brought forth. Without scalar, rotational invariance, the shape context (SC) algorithm has high time complexity and poor accuracy. In order to solve the above problems, a new shape feature descriptor-centroid distance ratio (CdR) is put forward. A better retrieval performance is achieved through the fusion of CdR and SC, and retrieval efficiency is enhanced by OpenMP.(2) Averification code(Completely Automated Public Turing test to tell Computers and Humans Apart,CAPTCHA) recognition method is proposed. First, CAPTCHA is pre-recognized by comparison of the shape context feature similarity. Then, more accurate recognition is conducted for several fallible characters through template matching. The experimentaltests show that this method has a good performance both on accuracy and efficiency.(3) A sketch editing and retrieval system is built. The system includes five functional modules:digital 2D creation, digital 3D creation, standard image editing, 3D model retrieval and video retrieval. The contourextraction process of hand sketch is elaborated. The sketch-basedimage retrieval and children’s cognition-oriented video retrieval are implemented. Finally, experimental results and analysis are exhibited.