Research And Implementation Of Sketch Skeleton Feature Retrieval Algorithm

The rapid development of science and technology has laid a solid foundation for the development of Internet and multimedia technology. With the enhancement of the performance of the handheld terminal equipment, as well as the appearance of the camera function, the number of images on the Internet has increased dramatically. It is very difficult for people to retrieve the desired image on the Internet. In order to solve this problem, many scholars at home and abroad will focus on this direction, and achieved remarkable results.The appearance of the touch function of the handheld device provides the condition for the convenience of the sketch retrieval. This new way of human-computer interaction has brought the new sense to the user. Compared to the previous PC side of the picture is not convenient, in the touch screen will be the user’s brain images drawn out on the image is very simple.But for different people, the background and cognitive is different, so will occur for the same image, different people will appear different cognitive outcome and lead to draw out the image differences, this sketch retrieval has brought great difficulties.And for natural images, can not be directly compared to the sketch. After extracting the outline of the natural image, the feature extraction is needed in order to realize the natural image retrieval.The effect of sketch retrieval is largely dependent on the accuracy of the feature description algorithm, and the correctness of the contour extraction. These are the difficulties of the sketch retrieval. This paper will sketch the sketch retrieval technology in detail, in-depth study. The main contents are as follows:(1)This paper describes the background and significance of freehand sketch retrieval, introduces the domestic and foreign research status of freehand sketch retrieval, and describes some of these algorithms. Based on the basis of these algorithms, and would compare the advantages and disadvantages of the algorithm proposed is used in this paper will feature extraction of skeleton and the minimum outer connected quadrilateral combined feature extraction algorithm and the good result was obtained in the experiment, verify the effectiveness of the algorithm.(2) In this paper, the algorithm based on the proposed a sketch detection system,the system will sketch to extract local features, through the bow bags(of words)model, the local feature transformation to global features to solve for the large-scaledatabase, the local feature too much and result in the curse of dimensionality problem.(3)In the retrieval process, will constitute the global features similarity comparison, according to the features of the alignment algorithm to derive the similarity, of the sort finally returned with the user input sketch the closest natural images.