Zhangkai Ni1, Lin Ma2, Huanqiang Zeng1, Canhui Cai1, and Kai-Kuang Ma3
1School of Information Science and Engineering, Huaqiao University, Fujian, China
2Huawei Noah's Ark Lab, Hong Kong
3School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore
In this letter, we make the first attempt to explore the usage of the gradient direction on the perceptual quality assessment of the screen content image (SCI). Specifically, the proposed approach firstly extracts the gradient direction based on the local information of the image gradient magnitude, which not only preserves gradient direction consistency in local regions, but also demonstrates sensitivities to the distortions introduced to the SCI. A deviation based pooling strategy is subsequently utilized to generate the corresponding image quality index. Moreover, we investigate and demonstrate the complementary behaviors of the gradient direction and magnitude for SCI quality assessment. By jointly considering them together, our proposed SCI quality metric outperforms the state-of-the-art quality metrics in terms of correlation with human visual system perception.
The contributions of this work:
Fig. 2. The SCIs and the corresponding direction distributions.
Fig. 1. The framework of our proposed IQA for SCI.
The framework of our proposed IQA for SCI is illustrated in Fig. 1. Our proposed IQA firstly computes the gradient magnitude and direction of the reference and distorted SCI, respectively. The deviation pooling is then used to pool the magnitude and direction similarities together to a final quality score.
Gradient direction computation:
The gradient direction map is calculated in terms of twelve directions from 0 to π by individually convolving the SCI with different convolution kernel at different directions. After obtaining the gradient direction maps for all twelve directions, the final gradient direction map is generated by selecting the maximum value among the responses overall directions. Fig. 2 demonstrates an example of SCI with Gaussian blurring, motion blurring, contrast change. The blue bars denote the gradient direction distributions of the corresponding SCIs, while the red bars indicate the difference between the original and distorted SCIs.
PLCC, SROCC and RMSE of IQA metrics on SIQAD database
GSS SSIM MSSIM IWSSIM VIF IFC VSNR MAD FSIM GSIM GMSD VSI SIQAD
If you have any questions, please feel free to contact Dr. Zhangkai Ni (eezkni@gmail.com).
Last update: Aug. 6, 2016
