Image Retrieval Based On LTP And SIFT Algorithm

With the rapid development of information age, the image database become very confused. The traditional text-based Image Retrieval technology is based on the image information annotation to implement retrieval, which has limitations and creates a subjective difference and large workload. This thesis do research into the algorithm for the image texture features on the basis of the Contented-based Image Retrieval technology. The main research work in this thesis is as follows:Firstly, this thesis analyzes and discusses on image retrieval based on LTP algorithm. Threshold in LTP can balance the illumination change, thus it greatly enhances the sensitivity of noise. The LTP algorithm has a good image retrieval performance and effectiveness under complicated factors such as image brightness, il umination, verified by experiments.Secondly, the thesis deeply researches the basic ideas and implementation of SIFT algorithm. It is proved that the SIFT algorithm for scaling, rotation, brightness change keeps invariance and for perspective changes maintains a degree of stability through a series of experiments.Finally, this thesis analyzes and discusses on image retrieval based on LTP and SIFT algorithm for the reason that using a single algorithm for image retrieval is difficult to achieve ideal effects. Using SIFT algorithm to detect image point at first, locate key point accurately, and then calculate the key neighborhood LTP feature with key point as the center, thus key point vectors can be obtained. Through the experimental analysis of the image retrieval system based on LTP and SIFT algorithm, it cannot only improve the image retrieval performance under the condition that image texture feature is not obvious, but also solve image matching problem in low il umination change.The thesis accomplishes the image retrieval system based on LTP and SIFT algorithm, through GUI retrieval interface built by software MATLAB2010 a, and the system achieves a good performance by a great deal of simulation experiments.