Sunteți pe pagina 1din 11

A Framework for Enhancing Depth Perception in Computer Graphics

EUNICE LEE 301179758


APR 2014
Depth Enhancem
ent in 3D CG
IAT 455 D100
TOPIC AND MOTIVATION OF AUTHORS
The 3D rendering methods and display technologies rapid development in the 3D technology
brought out the problem on the effectiveness of the usage of 3D. In the development of sys-
tem that enhance the depth perception of an input 3D scene, there needed to be an algorithm
that combines different depth cues and rendering mehtods automatically.
Currently, there are a number of methods that proposed to improve depth perception in 3D ren-
dering, but the methods are generally operating on specic domains or provide insufcient
solution in unifying the different depth enhancement methods. So, the authors of this study
examined the depth cues, the cue combination models and methods to present an algorithm
that automatically selects the proper depth enhancement methods for a particular scene.
WHAT WORK DOES THE STUDY BUILD ON
The study builds onto the previous researches that have investigated the interaction of different cues. And it extends upon the
methods for enhancing the depth perception in a 3D scene. They are proposing a new algorithm that automatically decide on
the depth cues using fuzzy logic and selecting the rendering methods that provides the cues based on the Knapsack problem.
As an extension of previous works, they considered the depth framework factors such as target tasks, spacial layout of the
scene and the costs of the rendering methods, and also tested their system using objective and subjective experiments,
BRIEF SUMMARY
Depth cues(aka visual cues) help the human visual system to perceive the
spatial relationships between the objects, construct the core part of depth
perception, and categorized as pictoral, oculomotor, binocular, and motion-
related cues There are several methods for enhancing 3D depth perception
according to the perspective-based cues, focus related cues, shading and
shadows cues, binocular and oculomotor cues, motion related cues and etc.
The suggested framework uses automatic selection of the proper depth cues
and rendering methods for the given scene, considers factors such as cue
averaging, cue specialization, range extension models of cue combination, the
target tasks,spatial layout of the scene and the costs of rendering methods
STAGE 1
The system stores a cue priority values. Fuzzy logic is used to calculate the cue priorities to combine different depth cues depending on factors
such as: task, distance of objects and scene attributes. Fuzzy logic based algorithm automatically determines the proper depth cues for the given
scene and task, it has been used to successfully model complex systems such as human intelligence, perception, and cognition, and provides a ro-
bust solution for the multi-input system whose mathematical modelling tis difcult
STAGE 2
The system selects the suitable rendering methods that provide the cues which are determined as high priority.It consider the costs of the methods
and try to solve the cost and cue priority trade-off. It uses knapsack model to select proper depth enhancement methods, evaluate and model
the trade-off between the cost and prot of a depth enhancement method The following knapsack models 3 steps are repeated muliple times to
obtain accurate estimation in the nal frames per second(FPS).
Step 1) cost-prot analysis step: prot and cost values are assigned to each depth enhancement method;
2) eliminate additional cost by unselecting some methods that provide only the cues that are already provided by other methods;
3)checking the use of helpers
The system applys the rendering methods to the given scene and produce a rened scene with a better depth perception.Depth test is performed
from the lights point of view, the points that cant pass the depth test should be in shadow.
Fog: Final color of each pixel interpolated between the surface color and the fog color according to the fog factor dependent upon the distance from
the view point
Proximity luminance: changes the luminance of the objects according to their distance from the viewpoint to provide relative brightness and aerial
perspective cue.
1)convert RGB color value to HSL;
2)modify the luminance value according to the pixels distance;
3)convert modied color back to RGB
Boundary enhancement: enhance important edge using the depth buffer based method. Derivative of depth values is used to calculate the spatial
importance.
1)calculate spatial importance function ;
2)color contrast of the whole image is modied by adding the spatial importance value multiplied by a constant to each color channel
Face tracking: use Bulbul et al.s system to provide motion parallax. The face position is used in determining the viewpoint
Multi-view rendering: provide binocular and oculomotor cues obtained by 9-view 3D lenticular display. The rendered scenes from different view-
points are combined using the interlacing operations for the display
STAGE 3
DATA AND RESULTS
The study have conducted formal experiments to evaluate the effectiveness of the proposed algorithms for depth and shape judgement tasks.
Objective experiment: average RMS error of the auto selection system is only 3.1% which meant the system showed the best result.
Subjective experiment:
1)depth perception: participants evaluated in terms of the ease of understanding the relative distance btw the objects in the scene
2)Shape perception: participants evaluated in terms of perception of shapes (curvatures,convexity/concavity) of objects
As a result, the scenes which are enhanced using the algorithm have the highest scores among other test cases for both tasks with a statistically
signicant (p < 0.05) difference
SIGNIFICANCE OF THESE FINDINGS
The human visual system perceives spatial information through various depth cues which are categorized as
oculomotor and visual cues. Depth cues (e.g., occlusion, depth-of-focus, relative brightness, size and texture
gradient) can be used to make objects look more like 3D. The study presented a systematic analysis to create a
general 3D scene rendering system to automatically select the better depth and shape enhancing cues.Cipiloglu
et al. present a framework for enhancing depth perception in computer graphics and have studied methods for
enhancing depth perception and have done several experiments where they tested several well-known depth
cues to understand their inuence upon perceiving depth in virtual environments.It presented the importance
of such depth cues (and their interaction) for depth perception accuracy in computer generated scenes.
SCIENTIFIC QUESTIONS REMAINING
What cue conicts can happen when using this automatic system?
What are the effects of the animation when using this system?
REFERENCES
Cipiloglu, Z., Bulbul, A., & Capin, T. (2010, July). A framework for enhancing depth perception in computer graphics. In Proceedings of the 7th
Symposium on Applied Perception in Graphics and Visualization (pp. 141-148).ACM.DOI=10.1145/1836248.1836276

S-ar putea să vă placă și