Image Repeating Element Extraction Based On Simple Interaction

Image segmentation is an important content in the field of image understanding and computer vision.The scope of application is widely used in medicine,aerospace,graphics,robotics,etc.,and has been highly valued in theoretical research and practical applications.Image segmentation is the key step from image processing to image analysis.The quality of segmentation directly affects the deeper information processing.Therefore,image segmentation plays an important role in the whole image processing,and has an important influence on the next feature extraction.The repeated elements are indispensable roles in our lives.It is practical to automatically separate the repeated elements from the image,which can alleviate the repeated search work and treat all the same repeated elements as a whole.Separate the repeating elements directly.All current research requires users to provide information about repeated elements for matching in the whole picture.We propose a framework for more automated extraction of repeated elements,which can be used to quickly find duplicate elements.The main content of this paper is to quickly extract the repeated elements in the image through simple interaction.Given a color image with repeated elements,the user roughly selects the region where the repeated elements are located,and we can get accurate repeating elements from it,and Greatly in line with user expectations.We use superpixels to speed up the algorithm,and along this line all clusters are clustered again,in order to obtain coarse foreground pixel information from superpixels without semantic information.According to the nature of the repeating elements and the brush information of the user interaction,the similarity/similarity of the two superpixels,the color difference between the superpixels,the spatial distance and the distance from the line drawn by the user are measured from three aspects.The relationship determines the similarity between two superpixels.We evaluate each clustering result.The criterion is whether the clustered clusters match the interaction information,that is,whether the aggregated results are consistent with the user.Enter interactive prompts.According to these final determinations of the foreground part containing the repeated elements,for more complicated cases,sometimes another set or more sets of repeating elements will inevitably appear when the user draws,and our original algorithm cannot distinguish which group is the user.Expected,it will fall into chaos.Then,based on this,we use the similarity propagation algorithm of the information dissemination mechanism to classify the complex prospects and obtain each set of repeating elements.Based on the process of simple interaction,users can experience the results of splitting repeated elements at once,saving a lot of time and effort.