融合主客观属性的个性化图像美学评价与自动裁剪研究

随着物质生活水平的不断提高，人们对物品在视觉上审美需求越来越多。因此，研究符合人们视觉审美体验的图像美学评价与自动裁剪方法具有重要意义。但是由于人的视觉审美体验具有较强主观性，导致人与人之间存在不同的审美偏好，因而大众化图像美学评价与自动裁剪方法难以满足特定用户的个性化视觉审美需求。鉴于此，本项目拟提出融合主客观属性的个性化图像美学评价与自动裁剪方法。首先，为了探究用户的主观属性，通过用户在社交网络中对图像的审美偏好行为，提出基于多模态融合的用户性格特征预测模型；然后，通过融合用户主观属性（性格特征）和图像客观属性（美学属性）来研究针对用户的个性化图像美学评价方法；最后，基于用户对图像的个性化审美偏好，在图像中自动生成满足其个性化视觉审美需求的最佳裁剪。本项目的研究成果可以应用于摄影摄像、图形界面设计、艺术设计、市场营销和智能相册管理等方面，具有重要的理论研究意义和实际应用价值。

With the continuous improvement of material living standards, people have a growing number of visual aesthetic demands for objects. Therefore, it is of great significance to study image aesthetics assessment and automatic cropping metrics that conform to people's visual aesthetic experience. However, due to the strong subjectivity of people's visual aesthetic experience, there are different aesthetic preferences among people, so generic image aesthetics assessment and automatic cropping metrics are difficult to meet the personalized visual aesthetic demands of specific users. Because of this, this project intends to propose a personalized image aesthetic assessment and automatic cropping method based on the fusion of subjective and objective attributes. Firstly, to explore the subjective attributes of users, the prediction model of users' personality traits based on multimodal fusion is proposed through users' aesthetic preference behavior for images in social networks. Then, the personalized image aesthetic assessment method for users is studied by fusing the subjective attributes (personality traits) of users and the objective attributes (aesthetic attributes) of images. Finally, based on users' personalized aesthetic preferences for images, the best cropping that meets their personalized visual aesthetic demands is automatically generated in the image. The research results of this project can be applied to photography and videography, graphic interface design, art design, marketing, and smart album management, etc., which have great significance in the field of both theoretical research and practical application.