Design And Implementation Of Image Matting System Based On Composition Aesthetic Evalutation

Image matting technology is a technical means to enhance the beauty of images.It analyzes the spatial position relationship of multiple subjects in the foreground and background,so as to achieve the purpose of removing irrelevant objects.There is a contradiction between the general demand for improving the beauty of images and the current high threshold for the application of matting technology,which requires users to have professional photographic literacy.Most of the existing research on matting technology focuses on image matting and image completion.The difficulty of the former is determining the position of unaltered subjects in the image,but also ensuring that the multi-subject spatial relationship after matting is optimized;the main difficulty lies in how to naturally and effectively complement the extracted areas,so as to enhance the beauty of the edited image.There’s also a lack of automatic image matting software,which combines the above-mentioned various technologies to achieve transparency and effectively meet user needs.In order to analyze the above problems,this paper first disassembles the matting technology into three main problems.First,how to perceive the beauty of the image,especially the analysis of the beauty of the composition that is closely related to the spatial position relationship,so that the system has the possibility of automatic image matting;Second,how to find the optimal solution from what is the target of extraction.After the beauty of the image can be judged,the next step is to optimize the image in some way.For the user,it is possible to directly identify the image in the image.Each object and process it,but for the machine,how to identify the operable objects in the image is also a problem that needs to be solved;Third,how to achieve a natural completion effect,complete the area to be completed according to the image information outside the cutout area,and obtain as real and reasonable background information as possible is also a problem that needs to be solved.In order to solve the above problems,this paper mainly conducts research based on the following aspects:1)This paper designs a composition aesthetic assessment algorithm based on subtask assistance.Aiming at the problems of low accuracy and poor interpretability of the existing methods,this paper firstly constructs a batch of datasets that can be used to drive the algorithm.Several test results show that the method in this paper can effectively perceive the beauty of the composition.2)This paper proposes a scientific image matting process.Key extraction and completion algorithms are selected and reproduced,and then the algorithm and process are optimized based on the analysis of the aesthetics of the image composition,so that the extraction target is reasonable and meets the user’s requirements,and that the complete effect is achievable by the extraction of the image and the completion of the image.Natural and enhanced beauty is the key to the extraction process.3)This paper develops an automatic matting system,which is based on the above designed and optimized algorithms and the automated mapcutting process,an automatic map-cutting system that can assist users in automatically improving the beauty of an image as it is constructed.The technical implementation details of the whole process are transparent to users.The system also provides user-defined cutouts and online training for online learning,which further assists users to obtain results that meet their own needs.