Datasets
Standard Dataset
IST - General Perspective Projection Dataset
- Citation Author(s):
- Submitted by:
- FALAH RAHIM
- Last updated:
- Wed, 10/26/2022 - 06:28
- DOI:
- 10.21227/7g1s-bt57
- License:
89 Views- Categories:
- Keywords:
Abstract
The subjective assessment of the generalized perspective projection (parameterized with projection center d) was conducted using eight omnidirectional images, in equirectangular format, taken from the Salient360! dataset [1]. The Stimulus Comparison Adjectival Categorical Judgment (SCACJ) was selected as an evaluation method. More details about generalized perspective projection, rendered viewport images, and the subjective assessment procedure can be found in [2].
[1] J. Gutiérrez, E. David, A. Coutrot, M. Perreira Da Silva, P. Le Callet, “Introducing UN Salient360! Benchmark: A platform for evaluating visual attention models for 360 contents”, International Conference on Quality of Multimedia Experience (QoMEX), Sardinia, Italy, May. 2018.
[2] F. Jabar, J. Ascenso, and M. P. P.Queluz, ‘Objective Assessment of Perceived Geometric Distortions in Viewport Rendering of 360° Images’, IEEE J. Sel. Top. Signal Process., vol. 14, no. 1, pp. 49–63, Jan. 2020.
- Three viewing directions were considered: front, 45⁰ to the right, 45⁰ to the left.
- For each viewing direction 10 viewports were rendered with d in {0, 0.25, 0.5, 0.75, 1} and FoV in {90⁰, 110⁰}.
- Viewports were rendered with d not equal to 0, evaluated as a pair with rectilinear projection (d = 0) as a reference.
- Score: +1 (slightly better) to +3 (much better) to the most pleasant viewport or 0 in case no difference was detected.
- Comparative mean opinion score (CMOS) was computed per viewport.
[2] F. Jabar, J. Ascenso, and M. P. P.Queluz, ‘Objective Assessment of Perceived Geometric Distortions in Viewport Rendering of 360° Images’, IEEE J. Sel. Top. Signal Process., vol. 14, no. 1, pp. 49–63, Jan. 2020.
