Title: ImageSpace Dynamic Transparency for Improved Object Discovery in 3D Environments
1Image-Space Dynamic Transparency for Improved
Object Discovery in 3D Environments
orBecoming Superman
- Ulf Assarsson ltuffe_at_ce.chalmers.segt
- Niklas Elmqvist ltelm_at_cs.chalmers.segt
- Philippas Tsigas lttsigas_at_cs.chalmers.segt
Winter Meeting 2006 Jan 11-13, Alingsås
2Why Become Superman?
- What if we could endow all human users with
Superman-like powers of observation? - Difficult in the real world
- Possible in the computer world
- Idea Give the users super-human vision
- See through walls
- See things far away
- See things too small to see with the naked eye
- Discover visualizations anew
3Example Supermans X-Ray Vision
"Where we come from everyone has see-through
vision, extra-strength and extra-speed!S No.
65/3 "Three Supermen from Krypton!
4On Superhero X-Ray Vision
Today's Superman possesses a wide range of
optical super-powers, including X-ray vision,
which enables him to see through all substances
except lead telescopic vision, which enables him
to focus on objects millions of miles away
super-vision, a combination of X-ray vision and
telescopic vision, which enables him to perform
such optical feats as peering through the wall of
a house thousands of miles away micro-scopic
vision, which enables him to examine the tiniest
atomic particles
- Sources Supermanica (supermanica.info) and
theSuperman Encyclopaedia (theages.superman.ws/En
cyclopaedia/) - Major components
- X-ray vision see through all substances and
materials except lead - Telescopic vision see (very) distant objects
- Supervision combination of x-ray and telescopic
vision - Microscopic vision see on a microscopic scale
5Benefits
- Superhero vision has a very important benefit
- Avoids problems with visibility and legibility in
3D environments - Can easily pinpoint important targets despite
occluding distractors - Main stumbling block of 3D information
visualization - Does not appear in 2D visualization
- Objects that do not intersect do not occlude each
other - Caused by the nature of the human vision system
- (But not the superhuman vision system...?)
- Example Cant see the forest for the trees.
6Problem Occlusion in 3D Environments
- Major problem inter-object occlusion
- Affects visual tasks related to object retrieval
- Object discovery finding all info-bearing
objects - Governed by object visibility
- Affects correctness may miss some information
completely - Object access retrieving info from objects
- Governed by object legibility
- Affects productivity effort needed to retrieve
information - Properties of the environment determines severity
of the occlusion effects - Object density, interaction and complexity
7Example Occlusion
8Image-Space Dynamic Transparency
- Idea Adjust transparency of surfaces to make
targets visible through occluding distractors - Observation The image space is perfect for
detecting instances of occluded targets and
dynamically adjusting transparency to allow the
user to see through surfaces - Can employ fragment and vertex shader
capabilities of modern programmable graphics
hardware - Achieve Superman-like cutaway effect of
surfaces to retain depth cues and spatial
information - In essence, we want to implement Supermans
X-ray vision in our visualization - Need to define a consistent model for dynamic
transparency
9Dynamic Transparency Model
- We define our model for dynamic transparency as a
set of rules - R1 All important objects (targets) in a scene
should be visible from any given viewpoint - R2 Objects are made visible by changing the
transparency level of occluding surfaces from
opaque (? 100) to transparent (? ?t gt 0)
within a cutout area enclosing the object - R3 Some surfaces are impenetrable and will never
be made transparent (cf lead for Superman) - R4 Objects are allowed to self-occlude
themselves - Cutout area convex hull (circle) or outline with
a gradient transparency border
10Example Russian Dolls
Normal vision
11Dynamic Transparency Model (2)
- There are some additional aspects on dynamic
transparency worth discussing further - Operational mode
- Active the user controls a searchlight on the
image space that reveals underlying occluded
objects - Passive all important objects are revealed at
all times (possibly using a selection filter) - See-through layer control
- Standard dynamic transparency peels off all
necessary surfaces to make an object visible - Sometimes you want to control how many layers
should be peeled away (search depth) - Example one-layer depth technical illustrations
Diepstraten et al. 2002 and 2003
12Algorithm Description
- Algorithm that fulfills requirements and model
- Multiple rendering passes
- Sort objects in descending depth (from the view)
- Initialize framebuffer with alpha 1.0 every
frame - Render differently depending on target or
distractor - Distractor Render to depth and then color
buffer, but blend using the alpha value of the
framebuffer - Target Render to depth and then color buffer.
Update the framebuffer alpha values using an
alpha mask - Alpha mask is multiplicatively blended to the
alpha channel of the frame buffer - Frame buffer becomes a cumulative alpha mask
- Alpha mask rendering is of special interest
13Alpha Mask Rendering
- Render object to mask buffer (256x256 texture)
- Iteratively grow the smooth alpha gradient
frame around the object - Idea Add a single line of increasing alpha to
the pixels surrounding the previous pass
Target object
14Implementation
- We have implemented this algorithm in C/C
- OpenGL for 3D graphics
- OpenGL ARB shader extensions
- Uses GLSL (GL Shading Language) for programmable
fragment and vertex shaders - Prototype implementation allows for the
construction of arbitrarily complex 3D scenes of
targets and distractors - Impenetrable surfaces can be omitted from the
method using the stencil buffer - Excellent framerates on the implementation (180
FPS on NVidia Geforce 6800) - Only screenshots, currently no real-time demo
- Lack laptop hardware capable of PS 3.0
15Results Screenshots
16Screenshots (2)
17Results Usability
- Plans are underway for a formal user study of the
method - Idea Present the user with a number of 3D scenes
of increasing complexity. Ask the user to search
for certain targets in a set of distractors. - Measure performance and correctness using dynamic
transparency versus standard 3D camera controls - Intuition (Much) better results using our new
method
18Conclusions
- Superhero vision has an important benefit
- Avoids visibility and legibility problems by
allowing for occluding surfaces to be made
(semi-)transparent - Our model for dynamic transparency supports this
mechanism in visualization applications - Targets are always visible through semi-opaque
cutouts in occluding distractors - Image-space computer graphics algorithm and
implementation of this model - Based on modern fragment shader programmability
- Promising results invite further research
- Including visual, performance, and usability
results
19Future Work
- Formal user study to measure efficiency
- Cutout geometry
- Interest-based dynamic transparency
- Importance/interest scale from 0,1
- Automatically derive object importance
- Technical illustrations
- Additional Superman vision techniques?
- Microscopic telescopic
20Questions?
- Contact informationNiklas Elmqvistelm_at_cs.chalme
rs.sePhone 46 31 772 1024Fax 46 31 772 3663 - Dynamic Transparency websitehttp//www.cs.chalme
rs.se/elm/projects/dyntrans/