Abstract
Methods for image Super Resolution (SR) have started to benefit from the development of perceptual quality predictors that are designed for super resolved images. However, extensive cross dataset validation studies have not yet been performed on Image Quality Assessment (IQA) for super resolved images. Moreover, powerful natural scene statistics-based approaches for IQA have not yet been studied for SR. To address these issues, we introduced a new dataset of super-resolved images with associated human quality scores. The dataset is based on the existing SupER dataset, which contains real low-resolution images. This new dataset also has 7 SR algorithms at three magnification scales. We selected optimal quality aware features to create two no-reference, (NR) opinion-distortion unaware (ODU) IQA models. Using the same set of selected features, we also implemented two NR-IQA opinion/distortion aware (ODA) models. The selection process identified paired-product (PP) features and those derived from discrete cosine transform coefficients (DCT) as the most relevant for the quality prediction of SR images. We conducted cross dataset validation for several state-of-the-art quality algorithms in four datasets, including our new dataset. The conducted experiments indicate that our models achieved better than state-of-the-art performance among the NR-IQA metrics. Our NR-IQA source code and the dataset are available at https://github.com/juanpaberon/IQA_SR.
Highlights
Image super-resolution (SR) refers to the construction of a high-quality high-resolution (HR) image from multiple or a single low-resolution (LR) input
Two sets of 45 and 73 perceptual quality-aware features were selected from a group of 306 features to create two NR-image quality assessment (IQA) metrics based on the working principle of NIQE [33] and IL-NIQE [34]
IL-NIQE and NIQE are models built on feature sets that have been selected for distortions which are different to the impairments possibly presented in the outcomes of SR algorithms
Summary
Image super-resolution (SR) refers to the construction of a high-quality high-resolution (HR) image from multiple (multiple-frame SR) or a single (single-image SR) low-resolution (LR) input. MFSR methods fuse frames with relative motion via interpolation [10], [11], iterative reconstruction [12], [13] and deep learning [14]. Previous works [16] have shown that PSNR and SSIM do not accurately predict perception of super-resolved image quality Other models such as the information fidelity criterion (IFC) [17] correlate better with human perception when evaluating super resolved images. These algorithms are full reference (FR), image quality assessment and require an original pristine image, which can be impossible to obtain.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
