A Hybrid Approach to Food Image Aesthetic Assessment via Structural and Color-Based Feature Integration

Assessing the aesthetic quality of food images requires capturing both structural and color-related attributes, a task that poses significant challenges. This study introduces a hybrid approach that integrates deep features of the VGG-16 model with RGB color histograms, providing a representation of the aesthetics of food image. Deep features capture structural and textural details, while RGB histograms highlight color-related attributes such as brightness, harmony, and balance. Our experimental results on a public dataset demonstrate that the proposed approach achieves state-of-the-art performance, surpassing baseline methods. The method maintains a balance between precision and recall across aesthetic labels, ensuring robust evaluation. Furthermore, the study underscores the complementary role of RGB color histograms in enriching deep feature representations, showcasing their synergy in enhancing overall performance. The effectiveness of SVM in managing high-dimensional feature spaces is also validated, reinforcing its suitability for aesthetic assessment tasks.