2017 Fiscal Year Research-status Report
Modeling the Perceptual Underpinnings for Quality Assessment of Restored Textures
| Project/Area Number |
17K00232
|
|---|
| Research Institution | Shizuoka University |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
大橋 剛介 静岡大学, 工学部, 教授 (80293603)
|
|---|
| Project Period (FY) |
2017-04-01 – 2020-03-31
|
| Keywords | quality assessment / restoration / enhancement / compression / visual perception |
|---|
| Outline of Annual Research Achievements |
The objective of this research project is to investigate the perceptual foundations of restored textures, to research and develop associated computational models and QA algorithms for restoration applications. In Year 1, the objective was to create a restored-image database containing quality scores for degraded textures and various restored versions of those textures. This database has been created.
In particular, the images were acquired, the textures were identified and segmented, and restorative patterns were synthesized. The restorative patterns were then added to HEVC-distorted versions of images to obtain two types of data relating to quality: (1) contrast thresholds for detecting the restorative patterns, and (2) optimal contrast scaling factors and quality improvement ratings. These data will provide important insights into how the human visual system judges texture quality.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
The research is largely on-schedule. Our main objective was to create the restored-image database, which includes not only the images and textures, but also the optimal scaling factors and the associated quality scores. In the original proposal, we did not suggest measuring the contrast detection thresholds. However, in order to model/predict the optimal scaling factors, a first step is to measure the visibility of the restorative patterns. Thus, we also measured the contrast detection thresholds. This decision to explore the visibility aspect limited the number of restorative pattern types that we were able to test. However, we are currently testing other types of restorative patterns both in terms of quality and detection thresholds, which will ultimately lead to more and better data.
|
| Strategy for Future Research Activity |
We have the added task of measuring both optimal scaling factors and contrast detection thresholds for the other types of restorative patterns. The advantage, however, is that we expect the contrast detection thresholds to be an important factor in modeling the visual impact of the restorative patterns. As stated in the original proposal, Year 2 will focus on model development. We have currently begun to develop such models, beginning with an investigation of the relationship between the visibility and optimal scaling factors. We believe that a perceptual model which uses basic similarity measures between the original vs. restored textures, will be able to quantify both the visibility and amount of quality change. One new M.S. student will focus exclusively on this modeling effort.
|
| Causes of Carryover |
The ordering of machines and desks was postponed due to additional students joining our research lab mid-year or later. For this reason, we have requested the remaining funds from Year 1 to be transferred to Year 2. Other students will contribute to this project by exploring other related perceptual aspects (e.g., other forms of distortion such as blur caused by resizing, and other texture-analysis measures such as texture regularity). These transferred funds will be used to pay for workstations/desks for these students. Some of the funds may be used for dissemination of the contrast-threshold study at a conference (currently under review).
|
