Aesthetic quality assessment of photographic images

Automatically assessing the quality of a photo is an important and challenging topic in visual computing. Previous methods mainly focus on image structures and image degradations caused by noise, distortion, or algorithmic operations such as digitization and compression. However, with the development and popularization of digital image capture devices, particularly mobile devices such as smart phones and pads, people pay more attention to the aesthetic quality of photographic images. In this paper, we propose an approach to assess the aesthetic quality of an image by computing scores for a set of fundamental and meaningful aesthetic attributes, namely clarity, contrast, and saturation. We firstly propose an adaptive window based clarity evaluation method, which performs more accurate evaluation for low depth of field images as well as normal images. Then, a method for contrast evaluation using histogram analysis is proposed. Furthermore, with the observation of a correlation between color intensity and saturation, we use logistic regression to estimate the optimal saturation of an image, and then a saturation rating is computed based on the difference between the optimal and real saturations. Finally, the aesthetic quality of an image is comprehensively evaluated as a weighted average between the scores of these three attributes. Experimental results show that our method performs well for evaluating the aesthetic quality of photographic images, which is consistent with the human visual aesthetic perception.