Abstract | ||
---|---|---|
Recently, an increasing number of tone-mapping operators (TMOs) have been proposed in order to display high dynamic nge (HDR) images on low dynamic range (LDR) devices. Developing perceptually consistent image quality assessment (QA) measures for TMO is highly desired because traditional LDR based IQA methods cannot support the cross dynamic range quality comparison. In this paper, a novel objective quality assessment method is proposed on the basis of sparse-domain representation, which has been well advocated as a powerful tool in describing natural sparse signals with the over-complete dictionary. Specifically, two indices, incorporating both local and global features extracted from sparsely represented coefficients, are introduced to simulate the human visual system (HVS) characteristics on HDR images. The local feature measures the sparse-domain similarity between the pristine HDR and tone-mapped L R images by leveraging the intrinsic structure with sparse coding. On the other hand, benefiting from the natural scene statistics (NSS), the global features are recovered from the sparse coefficients to account for the natural behaviors of tone-mapped images. Combining the local sparse-domain similarity and the global naturalness prior, validations on the public database show that the proposed sparse-domain model for tone-mapped images (SMTI) provides accurate predictions on the human perception of tone-mapped images. © 2016 IEEE. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1109/ISCAS.2016.7539023 | Proceedings - IEEE International Symposium on Circuits and Systems |
Keywords | Field | DocType |
High dynamic range, image quality assessment, tone-mapping operators, sparse representation | Kernel (linear algebra),Computer vision,Dynamic range,Pattern recognition,Human visual system model,Neural coding,Computer science,Naturalness,Sparse approximation,Image quality,Scene statistics,Artificial intelligence | Conference |
Volume | ISSN | Citations |
2016-July | 0271-4302 | 4 |
PageRank | References | Authors |
0.42 | 13 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
Lijuan Xie | 1 | 9 | 4.01 |
Xiang Zhang | 2 | 88 | 12.61 |
Shiqi Wang | 3 | 1281 | 120.37 |
Xinfeng Zhang | 4 | 195 | 12.61 |
Siwei Ma | 5 | 2229 | 203.42 |