IBR: View Interpolation Philipp Slusallek .

Uploaded on:
Picture Based Rendering. . . . . Geometry-based. Picture based. . Display/Environment Map. . Sprite/Imposter. Fraud/Sprite with Depth. . . LDI. Geometry. . View-dependentTexture Mapping. Lightfield/Lumigraph. View Interpolation. 1. Inserting between reach pictures - Chen and Williams, View Interpolation, 1993-Shade et al., Layered Depth Images, 19982. Correspondences and epipolar analys
Slide 1

IBR: View Interpolation Philipp Slusallek

Slide 2

Image-Based Rendering Lightfield/Lumigraph Sprite/Imposter LDI Geometry Imposter/Sprite with Depth View-subordinate Texture Mapping Panorama/Environment Map Geometry-based Image-based

Slide 3

View Interpolation 1. Interjecting between range pictures -Chen and Williams, View Interpolation, 1993 -Shade et al., Layered Depth Images, 1998 2. Correspondences and epipolar examination -Laveau and Faugeras, Collection of Images, 1994 -McMillan and Bishop, Plenoptic Modeling, 1995 3. Blend of view combination and transforming -Chen and Williams, 1993 -Seitz and Dyer, 1995  Require profundities or correspondences

Slide 4

Drawing RGBZ Images Various systems: 1. Need sorted splatting (Gauss, little rectangles, or focuses) -Chen and Williams 2. Need sorted tallness field -McMillan and Bishop 3. Micropolygon work -Marks et al. 4. Opposite beam following of stature field -McMillan 5. Crush and draw as a polygon work

Slide 5

View Interpolation Issues: Visibility/Occlusion Holes (Anti-)associating Shading (diffuse)

Slide 6

Directly from correspondences Laveau and Faugeras Algorithm p 2 p 1 Existing View2 Existing View1 p 3 l 13 l 23 New View 3

Slide 7

Merging RGBZ Images Problems: Overlaps -Oversampling -Occlusion Gaps or gaps -Undersampling -Expose concealed zones Chen and Williams 1993

Slide 8

View Interpolation [Chen/Williams\'93] Preprocessing: Correct correspondences for pixel (transform outline) into pieces (progressive) Obtain needs frame Z separating Display: Linear addition of transform guide Back-to-front rendering of squares Fill openings from adjoining pixels

Slide 9

Sprites/Imposters Purpose: Cache complex rendered geometry Decouple rendering overhauls from picture upgrades Drawing primitives Issues: Placement of polygon Warping of surface (point of view, relative) Validity of reserve Hierarchy Continuity with environment

Slide 10


Slide 11


Slide 12

Sprites/Imposters with Depth Better picture distorting: Wider scope of reuse Backward mapping just with homograph New mapping: Stored profundity delineate guide profundity outline (geometry) Backward mapping of shading utilizing profundity data d\' d\'

Slide 13

 I 1 d 1 ( I 2 ) I 2 Mapping with Depth Forward Mapping: Holes and associating

Slide 14

 I 1 ( I 2 ) d 2 I 2 Mapping with Depth Backward Mapping: What is d ?

Slide 15

 I 1 ( I 2 ) d 2 I 2 Mapping with Depth Solution: Forward guide profundity Reconstruct estimated geometry Backward guide shading

Slide 16

Layered Depth Images Idea: Handle disocclusion Store imperceptible geometry inside and out pictures Data structure: Per pixel rundown of profundity tests Per profundity test: RGBA Z Encoded: Normal course, remove Pack into reserve lines

Slide 17

Layered Depth Images Computation: Incremental distorting calculation Implicit requesting data Process in up to four quadrant Splat measure calculation Table query Fixed splat formats Clipping of LDIs

Slide 18

Layered Depth Images

Slide 19

Micro-Polygon rendering Goal: Fill Holes by expecting smooth geometry Approach: Connect neighboring pixels with polygons Delete polygons for substantial strides Render smaller scale polygons  Heavy load on representation framework

Slide 20

Some more Application: Stereo Example politeness L. McMillan, MIT

Slide 21

Some More Applications Holography Generating new perspectives of models Decoupling of rendering and picture overhauls Z is accessible Avoid antiquities Motion Capture Matching of model to picture information Morphing for adaptable models Video Rewrite/Talking Head

Slide 22

From Michael Cohen

Slide 23

From Michael Cohen

Slide 24

From Michael Cohen

Slide 25

Image-Based Rendering Limitations Static scene geometry Fixed lighting Constrained look-from or take a gander at point Possibilities: Complete demonstrating and rendering framework? Joining of Computer Vision and Computer Graphics?

View more...