Abstract
The role of image quality models was investigated to determine its existing limitations using 8 types of artificial and natural still monochrome pictures. The pictures were impaired by 6 coder relevant distortion of different strengths as subjects gave quality judgments that led to an image quality database. The mean standard deviation of the Mean Opinion Score of observer judgments were correlated by Mean Square Error (MSE), Root MSE and Peak Signal to Noise Ratio (PSNR). Results showed that raw error metrics improved the correlation with the observers' estimates, with the best result for PNSR. Filtered and masked as well as segmentation error metrices further revealed significant deviation from observer estimates.
| Original language | English |
|---|---|
| Pages (from-to) | 321-325 |
| Number of pages | 5 |
| Journal | IEE Conference Publication |
| Issue number | 465 I |
| DOIs | |
| State | Published - 1999 |
| Event | Proceedings of the 1999 7th International Conference on Image Processing and its Applications - Manchester, UK Duration: 13 Jul 1999 → 15 Jul 1999 |