| Auto object segmentation in unconstrained images has been widely applied in several areas of image processing, such as industrial application, modern medicine science, face recognition, traffic control and the like. Meanwhile, along with the rocketing development of the mobile Internet and smart terminals, taking and sharing cloth photo and matching clothes has become an indispensable habit in our daily life. Cloth segmentation and color extraction majored in cloth image has also become a popular research topic in this field.For this topic, in the process of dress collocation, cloth segmentation and color extraction are two key steps. The cloth segmentation algorithm, which is based on the graph theory, regards each pixel as a node in the graph. The RGBA distance value between pixels is regarded as the weight of the edge. As a consequence, the segmentation problem is can be solved by calculating the minimum cut of the graph. The interactive image segmentation algorithm involved the user interaction into this problem. GraphCut, as a typical instance of interactive image segmentation algorithm, calculates the energy function’s minimum value to get the result, and it can be applied in the cloth segmentation for its balance of efficiency and effectiveness. The main contribution of this paper is to use watershed method preprocessing the image, and conduct an light and convenient improved GraphCut to achieve the segmentation. Besides, this paper selects a clustering algorithm based on expectation maximization to take the image color extraction.Based on the cloth segmentation and color extraction result, users can match their clothes and share it. We depict an efficient system, which arrive at an eminently suitable point for the cloth segmentation. This paper chooses iOS as the implementation platform, and constructs the4-layer algorithm-based architecture. The experimental results demonstrate that we achieve an ideal effect in both cloth segmentation and color extraction. |
PDF Full Text Request