Research Of Image Retrieval Based On Curvelet Transform And SIFT-PCA Algorithm

Recent years have seen a rapid increase of the size of digital image collections. Both multimedia technology and digital technology generates a large number of multimedia data which include images, audio, video and other information. How to retrieve the desired content quickly and accurately from a large number of information has become a hotspot issue in the field of multimedia research. In order to solve the issue, people come up with CBIR (Content-based Image Retrieval).CBIR is the use of the nature of the image such as color, texture, shape, spatial relations, and image processing technology, pattern recognition and computer vision technology etc to achieve image retrieval. Due to the shape feature of image is better meet people’s visual perception and the users prefer to retrieve image with it. However, the realization of Shape-based Image Retrieval has a certain degree of difficulty because of the inherent specificity of shape feature. Therefore, Shape-based Image Retrieval, a challenging subject in CBIR, has very important significance of the research and future prospects.The paper firstly introduces the research background, status and the applied related technology based on the CBIR and Shape-based Image Retrieval, especially deeply investigate the crucial technology of Shape-based Image Retrieval. As some usual methods are always complex and need large computation. This paper proposes a novel method based on improved CT (Curvelet Transform) algorithm to overcome this drawback.Then for the limitations of the scale-invariant feature points extraction algorithm are discussed at present, including larger amount of calculation, more complex matching, and lower retrieval rate, this paper proposes an improved algorithm SIFT (Scale-invariant Feature Transform) with PCA (PCA, Principal Component Analysis) that is SIFT-PCA. This algorithm utilizes SIFT algorithm to extract feature vectors of the image feature points, and then transforms the feature vector space to another by the improved PCA algorithm, gains the most representative of the characteristic parameters to realize the feature point vector dimension reduction. The method guarantees the robustness of traditional sift algorithm, decreases the calculation amount, and enhances the real-time.Finally, in order to evaluate the retrieval performance of the paper proposed, we compare the proposed improved CT algorithm, SIFT-PCA algorithm, and the two features fused with the traditional feature extraction algorithm. Through the experiment shows the proposed method’s high efficiency and practical in image retrieval.Experiments are carried out in a standard Corel dataset, a Coil-100Color database, a standard MPEG-7image database, and a PI100commodity image database to test the efficiency and robustness of the proposed system. Experimental results show the proposed algorithm can retrieve images more efficiently than traditional learning algorithms.