Research Of Image Contour Detection And Its’ Application In Pornographic Image Recognition

Image is the main media of human communication and cognition, and it is the main source of information. Scientific research shows that people get about 80% of the information from the visual images. Image contour is the basic feature of the image, which contains a lot of meaningful information, is a kind of low-level features. The essence of image contour is to extract the meaningful image segmentation by using the algorithms, it is the premise or basis of image processing, and it has been a hotspot research field in image processing in the last several decades. Image contour detection has a direct impact on the process of image processing, so it is of far-reaching significance to find the boundary and inner contour of image in many computer tasks(such as image segmentation, object recognition, scene understanding, etc.).Because of much prominent boundary of vision, it is impossible to find a common boundary detector for all images. Therefore, image contour detection is still a challenging problem. With the convenience of the use of some public image datasets, different methods have been put forward,such as methods of local features, global features or sparse coding. The use of local features is insufficient in the details of the image contour, while the use of sparse coding method for image contour detection is able to detect contour details, on the other hand, it increases the number of non boundary points. In this paper, we propose a method which uses the combination of image local features and sparse code features to train our model. Then, we use the spatial and curvature information of the image pixels to ensure the accuracy of the contour detection and region consistency, then improves the average accuracy of image contour detection. Experimental results show that contours obtained by multi view clustering algorithm improves the contour accuracy and also ensures the smoothness of the contour. Compared with other algorithms in the public datasets, our proposed algorithm achieves comparable.In this paper, we also apply our image contour detection algorithm to the recognition of pornographic images, and conduct extensive experiments on the related datasets. The results show that the proposed algorithm can effectively remove the background of the image in a certain degree, thus improving the recognition rate of pornographic, which also indirectly shows the practicability of the algorithm.