Blind Image Quality Assessment Using Standardized NSS and Multi-pooled CNN

This paper proposes a blind image quality assessment (BIQA) method that combines the natural scene statistics (NSS) based feature extraction and the multi-pooled image feature extraction. The two features are concatenated and fully connected layers are utilized to output the image quality score. In the NSS feature extraction part, the mean subtracted contrast normalization is first conducted and then a CNN structure is followed. In the multi-pooled image feature extraction part, the structure of spatial pyramid pooling (SPP) is effectively embedded in a CNN structure. The proposed BIQA method is an end-to-end learning technique. In experiments, the performance of the proposed method is compared with that of the state-of-the-art methods through the use of several databases. The experimental results show a superior accuracy of the proposed method for BIQA.