Image quality assessment using a SVD-based structural projection

作者：

Highlights：

• Singular value decomposition is regarded as a structural projection transform on a set of under-complete bases.

• Image degradation can result in the changes of the projection bases and the coefficients of this structural projection.

• We use a block structural projection to detect the local image distortion feature vectors.

• Both perceptual spatial pooling and neural networks are employed to combine feature vectors to a single quality score.

摘要

•Singular value decomposition is regarded as a structural projection transform on a set of under-complete bases.•Image degradation can result in the changes of the projection bases and the coefficients of this structural projection.•We use a block structural projection to detect the local image distortion feature vectors.•Both perceptual spatial pooling and neural networks are employed to combine feature vectors to a single quality score.

论文关键词：Image quality assessment (IQA),Human visual system (HVS),Singular value decomposition (SVD),Structural projection,Neural networks (NN),Spatial pooling

论文评审过程：Received 15 October 2012, Revised 17 January 2014, Accepted 17 January 2014, Available online 4 February 2014.

论文官网地址：https://doi.org/10.1016/j.image.2014.01.007