Research On Image Aesthetic Evaluation Algorithm Based On Multi-index Fusion

In recent years,the application scenarios of character images have become more and more abundant.However,in the acquisition process,the differences in the aesthetic level and shooting environment of the photographer lead to the uneven aesthetic level of the character images.Therefore,it is of great practical significance to carry out the research on aesthetic quality evaluation and select valuable high aesthetic images from the massive character images.The aesthetic evaluation of character images has strong subjectivity and complexity,and there is no unified and perfect evaluation standard at present.Therefore,based on human visual characteristics and aesthetic habits,this paper puts forward a calculation method for evaluating the aesthetic value of character images to further improve the accuracy of aesthetic evaluation,realize the detailed analysis of various angles,and provide reference for the later image optimization.The main research contents of this paper are as follows :(1)Research on Aesthetic Feature Extraction Algorithm of Character ImageIn view of the lack of specific evaluation of the aesthetic feeling of the character image from different angles,which is not conducive to the subsequent targeted improvement and optimization,this paper determines the image characteristics that affect the aesthetic feeling of the character image according to previous studies and photographic rules,including technical characteristics(contrast characteristics,clarity characteristics,saturation characteristics),perception characteristics(composition characteristics,visual balance characteristics,depth of field characteristics)and social characteristics(eye feature,smile feature,intimacy feature),and quantify the above features to obtain the aesthetic feeling of the character image from different angles.The evaluation results of each feature are clear and intuitive,which can play a guiding role in improving the quality of image aesthetics,and lay the foundation for the overall evaluation of image aesthetics.(2)Establishment of evaluation model and experimental analysisAiming at the problem that the aesthetic evaluation results of the existing algorithms are not accurate enough to keep consistent with the real feelings of human beings,this paper proposes an aesthetic evaluation model suitable for single-person images and multi-person images by integrating the indicators in part(1).The test results on public data sets show that the model can achieve 98.08 % classification accuracy of high and low aesthetic senses,which is better than the other two algorithms with more references.It is proved that the evaluation model proposed in this paper can simulate human aesthetics to a large extent and realize effective evaluation of the aesthetic degree of character images.(3)Design and implementation of image aesthetic evaluation APPAiming at the problem of lacking the application of aesthetic evaluation algorithm for character images combined with actual scenes,this paper analyzes and disassembles the needs of young user groups who are willing to take pictures,designs and implements an Android mobile application that can evaluate and intelligently select character images from the aspects of interface and technical architecture.The algorithm model proposed in this paper is integrated into the mobile application to realize the algorithm landing.The test shows that the mobile phone application can meet the needs of users and has good performance.