Image Quality Assessment Scheme Based on Structural Contrast Index and Gradient Similarity
Image quality assessment (Anmin et al. in Image Proc, IEEE Trans 21(4):1500–1512, 2012 [1 ]) is very important for image processing. A good image evaluation algorithm is consistent with subjective evaluations and has low computational complexity. A lot of
- PDF / 149,578 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 100 Downloads / 224 Views
Abstract Image quality assessment (Anmin et al. in Image Proc, IEEE Trans 21 (4):1500–1512, 2012 [1]) is very important for image processing. A good image evaluation algorithm is consistent with subjective evaluations and has low computational complexity. A lot of image quality assessment methods have been proposed in recent years. Structural Contrast Index (SCI) has been proved can effectively reflect the complexity of image texture and model the masking effect of human visual system (HVS), so SCI is used as an important feature. HVS is very sensitive to edge region, however, SCI can’t correctly model the edge region structure. So the gradient similarity was incorporated into our method. An image quality assessment scheme based on structural contrast index and gradient similarity was proposed in our paper. Extensive experiments conducted on TID2013 image database demonstrate the performance this scheme is slightly better than the state-of-art methods not only on prediction accuracy but computational complexity.
Keywords Image quality assessment Masking effect Structural contrast index Gradient similarity
1 Introduction With the development of computer vision and image processing, a lot of image quality assessment methods are proposed. According to the availability of reference image, the presented metrics can be classified into full-reference ones, reduced-reference ones and no-reference ones. In this paper, the discussion focuses on FR methods. L. Liu Y. Zheng (&) W. Wang Faculty of Printing, Packing Engineering and Digital Media Technology, Xi’an University of Technology, Xi’an, China e-mail: [email protected] Y. Zheng Shaanxi Provincial Key Laboratory of Printing and Packaging Engineering, Xi’an, China © Springer Nature Singapore Pte Ltd. 2017 P. Zhao et al. (eds.), Advanced Graphic Communications and Media Technologies, Lecture Notes in Electrical Engineering 417, DOI 10.1007/978-981-10-3530-2_41
325
326
L. Liu et al.
IQAs like peak-to-noise ratio (PSNR) and root-mean-square error (RMSE), which are based on the mean square error (MSE) correlate poorly with HVS [1]. As we know more about the HVS, some HVS based methods are operated on image contrast rather than pixel value are presented. The Mean Structure Similarity Index (SSIM) proposed in paper [2] is quite a big breakthrough in image quality assessment modeling. Structure based methods like SSIM is based on the hypothesis that HVS is very sensitive to some structure information in the visual scene. So the result is more consistent with perceptual quality. The upgraded versions MS-SSIM [3], measure the structure loss of image. It is well known that edges are very important image structures and crucial for visual perception. Recently, gradient/edge based or gradient combined methods were proposed in succession. Multiscale visual gradient similarity (VGS) [4] is a very reliable one to some extent. Cheng et al. presented a GSD [5] method based on gradient magnitude and orientation. It has a good performance on LIVE image database and a
Data Loading...
