Realtime 3D Computer Graphics Virtual Reality

Transcription

1 Realtime 3D Computer Graphics Virtual Reality Human Visual Perception The human visual system 2 eyes Optic nerve: 1.5 million fibers per eye (each fiber is the axon from a neuron) 125 million rods (achromatic low light sight) most outside the fovea 6 million cones (high detail color sight) most concentrated inside fovea Radio 3 types: R, G, B sensitive to long, middle and short wavelengths (!) respectively 350 Approximately 100:1 compression from the number of receptors to the number of fibers in the optic nerve. Quick transmission rate through the optic nerve Monocular visual field is 160 (w) x 135 (h) Binocular visual field is 200 (w) x 135 (h) Processing is not uniform across the visual field: 25% of cortex is devoted to the central 5 of the field of view. Light! (nm) 780 Heat

2 Human visual perception Human visual perception processes position/orientation and movement in 3 dimensions plus color. The third dimension depth is processed based on several physiological and psychological depth cues. Depth cues 1 can be binocular or monocular. r-eye convergence c Binocular depth cues: Convergence Difference in the direction of the eyes. Our eyes point slightly inward for closer objects. Only effective on short distances (< 10 meters). Binocular Parallax Difference in the sensed images by our two eyes. Our eyes see the world from slightly different locations;! images sensed are slightly different. Human visual system is very sensitive to these differences;! most important depth cue for medium viewing distances. Can be used to achieve depth sense even if all other depth cues are removed. 1 depth cues following (Okoshi, 1976) r-eye different images (size, position and content) on the real retina as well as on the virtual projection plane. Human visual perception Monocular depth cues: Monocular Movement (motion) Parallax Depth perception by moving each of our eyes (head). Depth information is extracted from consecutive similar images in the same way as images from different eyes are combined. Retinal Image Size Brain compares the sensed size of an object to its known real size. t

3 Human visual perception Monocular depth cues continued: Linear Perspective Straight parallel lines meet in the horizon. Important depth cue. Texture Gradient Closer objects look more detailed. Objects with smooth surface textures are usually interpreted being farther away (especially true if the texture spans from near to far). Occlusion, Overlapping Out of sight blocking of objects. Aerial Perspective Distant objects (mountains in the horizon) look always slightly bluish or hazy due to small water and dust particles in the air between. Shades and Shadows Objects shadowing others are closer to light sources. Useful to resolve ambiguities. Bright objects seem to be closer to the observer than dark ones. (Example: Three dimensional looking WIMP interfaces.) Simulating visual stimuli 3D CG rendering provides: Methods: Retinal Image Size Linear perspective Texture Gradient Occlusion Aerial Perspective Shades and Shadows Perspective projection High tessellation, LOD, texturing (images, bump maps, normal maps, height maps ) Occlusion culling, z-buffer algorithm fogging, atmospheric models Special lighting equations, shadow maps, shadow casts VR requires immersion and hence a simulation of visual stimuli which provides a mature depth perception: Convergence Binocular Parallax Monocular and binocular Motion Parallax Stereoscopy, channel separation Head (motion) tracking, dynamic view frustum

4 Implementing additional depth cues Stereoscopy: Render from two offset eye points (IPD) or center of projections (COPs). off-axis projection " (and # in 3D): VPN " Right Motion eye parallax: p COS Image plane/screen fixed w.r.t. world p`` binocular parallax p` Right eye View Plane Normal (VPN) COP t-1 Track head (and hence eye movements) and calculate new perspective projection. translation Calculate dynamic view frustum in case of image plane fixed w.r.t. world. COP t p p q p t q t q t-1 p t-1 Image plane/screen fixed w.r.t. world p`` binocular parallax p` Center Of Screen (1) (COS) Image plane/screen 1 fixed w.r.t. (right) eye motion parallax nearness: leftness: Stereoscopy Features of binocular parallax Negative: object in front of screen Zero: object on the screen Positive: object behind the screen Focus vs. convergence Focus is on image plane Convergence is on virtual object! Large parallax puts strain on the eye. Stereoscopy methods Feed each channel and its rendered picture to one specific eye by using one screen per eye (HMD). time-multiplexing generated images (shutter glasses). filter images through polarization filters. filter images using color filters (anaglyph). using auto stereoscopic displays. Shutter Technology Close left eye when right eye image is displayed and vice versa. Controlled through infrared or wired up. Usually connects to V-sync signal (vertical retrace of CRT).

5 Stereoscopy Polarization Light: wave length and direction of polarization. Two components orthogonal to each other. normal light polarized light Filters can block certain directions of polarization. See through linear polarization (use two projectors): Left view: vertical filter in front of the lens. Right view: horizontal filter in front of the lens. Wear glasses with polarizing filters. : vertical Right eye: horizontal Stereoscopy Linear polarization Can t tilt head Little ghosting See through circular polarization (using two projectors): Left view: clockwise filter Right view: counter clockwise filter Allows arbitrary head orientations In general more ghosting than linear polarization linear polarization circular polarization Anaglyph stereo Combine each channel s R,G,B values by two complementing transformations to calculate an integrated channel. Several anaglyph version exist. Usually black/white images, color possible but filter and image colors may interfere. Example for a red/blue transformation:

6 Stereoscopy Pulfrich effect At low light levels the eye-brain visual response is slower. Using a neutral (transparent gray) filter over one eye. Movement perception by that eye will lag behind perception by the unimpeded eye. Lag induces a difference in the images perceived by each eye. This induces a binocular vision illusion of depth. Auto stereoscopic displays Holographic displays, e.g. laser projection on gas or fluids. Modified LCDs Assign alternating pixel columns for each eye. Filter outgoing light by prisms or by two vertically striped masks located in front of the LCD. Slightly dislocate the masks in depth and displace them horizontally. Autostereograms SeeReal Technologies C-nt res. 1600x1200 mono res. 800x1200 stereo sweetspot distance 650 mm sweetspot width/depth 50/150mm