Classification of Aesthetic Photographic Images Using SVM and KNN Classifiers

Aesthetics of photographic images is related to beauty appreciation and is considered vital research in the disciplines of image processing and computer vision. The aesthetic quality of images is a subjective matter and is very challenging to accomplish in order to reach beauty appreciation. Despite the absence of standardization for measuring aesthetic quality, we tried to offer a simple, yet precise, a method to express aesthetic beauty. Aesthetics of photographic images is highly desirable in many applications such as information retrieval applications and others. In this paper, we classify the features of aesthetic images-which are: contrast, noise, blur/focus, image composition, color saturation, image histogram and color harmony-using the Support Vector Machine (SVM) and k-nearest neighbor (KNN) classifiers. After the application of feature extraction and classification processes over images from a video dataset of Atomic Visual Action, classification of the images resulted in two classes: high-aesthetic- quality images and low-aesthetic-quality images. We found that both classifiers gave high accuracy rates, which were relatively close, SVM achieved 86.6667% success rate, while KNN outperformed it with 87.0968% accuracy. On the other hand, the F-score of the SVM was much higher, where SVM gave an F-score of 2, and KNN gave an F-score of 0.