Blind screen content image quality measurement based on sparse feature learning
- 148 Downloads
Recently, the perceived quality measurement of screen content images (SCIs) has become an active research topic. In this paper, a blind image quality measurement (IQM) metric for SCIs based on the learning of sparse features via dictionary learning is proposed. First, to extract the sparse features, histogram representations from multi-scale local gradient patterns are integrated to form a dictionary. Subsequently, using a pursuit algorithm, the sparse features of the distorted SCIs are efficiently coded by this dictionary. Finally, to obtain the final quality of the distorted SCIs, a machine learning algorithm is utilised to combine the sparse features into a final quality score. The results of extensive simulations conducted show that the proposed blind IQM metric consistently obtains competitive performance and is in line with human beings perceive.
KeywordsScreen content images Sparse features Multi-scale local gradient patterns Quality measurement
This work was supported by the National Natural Science Foundation of China (Grant Nos. 61502429, 61505176), the Zhejiang Provincial Natural Science Foundation of China (Grant No. LY18F020012), the Zhejiang Open Foundation of the Most Important Subjects, and the China Postdoctoral Science Foundation (Grant No. 2015M581932).
- 11.Zhang, Y., Chandler, D.M., Mou X.: Quality assessment of screen content images via convolutional-neural-network-based synthetic/natural segmentation. IEEE Trans. Image Process. (2018) https://doi.org/10.1109/TIP.2018.2851390