No-reference Image Quality Assessment Based on Ensemble Convolutional Neural Network
Qian Wan, Qingbing Sang · 2020
We propose a no-reference image quality assessment based on ensemble convolutional neural network. Firstly, the distorted image is cut into image patches, and the image patches are pre-processed by performing local contrast normalization. Then use convolutional neural networks to extract features of image patches, which eliminates the trouble of manually extracting features, four convolutional neural network models with different structures are designed. Finally ensemble the four network models to improve accuracy of image quality assessment. Experimental results on LIVE and CSIQ database show that the proposed method can predict the image quality well and has good generalization performance.