Digital painting image quality evaluation based on visual reasoning and multi-layer perception

With the popularization of digital painting, image quality evaluation has become a crucial topic. To improve the accuracy and efficiency of digital image quality evaluation, an innovative digital painting image evaluation model is proposed. The proposed method designs a parallel network structure that utilizes multi-layer perceptrons for deep feature extraction, while utilizing a visual geometry group network with 16 layers for shallow feature extraction. The final quality evaluation score is obtained by combining the two scores, which are separately scored for the extracted features using structural similarity and Euclidean distance. The experiment outcomes indicate that the raised feature extraction model achieves an accuracy of 0.846, which is better than the 0.782 and 0.749 of the two compared models. In terms of accuracy and recall, they are 0.842 and 0.681 respectively, both higher than the comparison model. Meanwhile, in the results of image quality evaluation indicators, the proposed model shows excellent evaluation performance. The proposed model fully demonstrates superior feature extraction ability and evaluation accuracy in the quality evaluation of digital painting images, providing reliable technical assistance for practitioners in the digital painting industry.