No-Reference Stereo Image Quality Assessment Based on Transfer Learning

In order to apply the deep learning stereo image quality evaluation, two problems need be solved: The first one is that we have a bit of training samples, another how input dimensional image’s left view or right view. this paper, transfer 2D evaluation model and method solves problem; use principal component analysis used fuse views into an in solve second problem. At same time, preprocessed by phase congruency transformation, which further improves performance algorithm. structure convolution neural network consists four layers three maximum pooling fully connected layers. experimental results on LIVE3D database show prediction score good agreement with subjective value.