Stereo – In Theory and in PracticeMartin Ericsson

3D?

● How do we perceive “3D” or depth in a picture?● 3D or stereo? (as in stereoscopic)● Content for today

– Depth Cues

– Theory

– Display Techniques

– Demos

Monocular Depth Cues

● Secondary depth cues● All depth cues help the user to form a mental

image of the data displayed● Occlusion● Perspective● Relative size● Motion parallax● Shading and shadow● Atmosphere● Details

Occlusion

● When two objects are at the same line of sight the object closest to the viewer should occlude the other object

● This gives information about the relative position of the two objects

Occlusion cont.

● Objects that the user have no prior information about is hard to approximate

● Relates to both (relative) size and distance● Use overlap to help viewer● Depth buffer in computer graphics

Relative Size

● As an object gets closer to the viewer it fills up more of our view space and by this looks larger

● So larger objects are perceived as being closer to us

● The type of object also makes a difference here, i.e., if the user has prior knowledge of the object

● Projection calculation in computer graphics● Also relates to perspective

Perspective

● Orthographic projection where two parallel lines never meet can be practical in some applications

● But to give depth feedback to users a perspective projection should be used

Perspective

● Put things on different depth to show inter relationship

● Lots of and details gives by perspective much depth cues

Detail

● The amount of detail an object have also gives a hint on how close it could be

● More detail means that an object is closer● Compare this to a landscape. By your feet you

can see grass and individual stones. A bit further away they are not visible anymore only bigger thing and at a distance it is all “a blur”

Detail cont.

● Much detail gives cue for closeness● Use texture maps for example to add details● Also geometric detail is important● Adds much spatial information● Also adds more motion parallax

Motion Parallax

● Object (edge) speed gives hint on depth● Closer object is perceived as moving faster than

distant objects● Compare to what you see outside the window

when riding on a train for example● Related to detail, lots of geometric detail gives

more motion parallax● That is that the users sees many edges moving

at different speed

Shading and shadow

● Shading gives a sense of curvature of an object● Shadows gives information about a relationship

between objects● One of the most important cues!

Shading and shadow

● More complex objects need these cues to make sense

● Help the user to form a mental image

Atmospheric scattering

● Light traveling through the atmosphere is scattered and absorbed

● This makes objects that are at a greater distance to the viewer to have more of a bluish tone.

Atmospheric scattering cont.

● This is often implemented as some kind of fog effect

● Related to this is also to add focal blur

An example

● Adding depth cues● Method

– Molecule model

– Imported a database file into Paraview (an application build upon VTK)

– Create geometry and export to Blender (www.blender.org)

– Render in Blender with different settings

Example (no shading)

Example (shading)

Example (shadows)

Example (ambient occlusion)

Example (edges = contrast)

Binocular depth cues

● Primary depth cues● All the secondary can we have in any picture● How come reality has more depth than a photo?● Three primary depth cues

– Accommodation

– Convergence

– Retinal disparity

Accommodation

● To focus on something we apply pressure to deform the eye lens

● The amount of pressure is a depth cue● Works with one healthy eye and only at closer

range ● This depth cue is not something that we have

equipment for today

Convergence

● When we look at something we rotate our eyes so the line of sight intersects at the point we are looking at

● The amount of rotation is also a depth cue for us

● Most people cannot diverge their eyes● We can use this with for example a stereoscopic

display

Retinal disparity

● Majority of us have two functional eyes.● We get two inputs to our brain but we only have

mental image.● The two images are compared and the

difference between them also gives a strong depth cues

● Stereopsis!● Also supported in a stereoscopic display

Display Techniques

● Four common techniques for single display plane– Polarization

– Anaglyph “Red-green”

– Time multiplexing

– Spectral Multiplexing

● Other types do exists, e.g., volumetric display, autostereoscopic displays etc.

Display Techniques cont.

● Holographic Optical Element

Display Techniques cont.

● Ground principle for stereoscopic image– Two images with a bit different point of view

– Make sure that “right” image reaches “right”

● Two ways to do this practically– Side by side rendering

● Compatible● Some side effect (later)

– Quad buffer● Needs special hardware

Display Techniques cont.

● Projections– Need a way to project the world onto a planar

display

Parallel Toe-in Off-axis

Display Techniques cont.

● Screen space vs. View space

Display Techniques cont.

● Screen space vs. View space

Display Techniques cont.

● Screen space vs. View space

Anaglyph

● “Red-green” stereo● Both sources are weaved into one● Filtered based on color channels● Works on computer screens, printed form etc.● Support in VTK

Anaglyph explained

Time multiplexing

● Uses time to weave images together● One eye is occluded while the other sees the

image.● Uses active glasses that are a bit heavier than

other techniques

Frame n Frame n+1

Polarizing

● Separating by polarization of light● Two sources with different polarized filters● Glasses with different polarized lenses● Requires a special screen that preserves

polarization

Spectral Multiplexing

● Technique used here is called INFITEC and it is spectral multiplexing

● The visible light spectra is divided into six parts, three for each eye

● Each eye gets half of the light in the blue range, half from the green and half from the red

Spectral multiplexing

Signal out

Left eye Right eye

Method

● Computer with two video out● A box that enhances the signal● Two projectors with filters

– Notice the shift in color

● Screen– Regular screen

● Glasses with filters– Light weight

Demonstration

● Questions?

Ewert Ingela Bo