Impacts of Viewing Geometry on Combination of Disparity and Texture Gradient Information

Effects of viewing geometry on combination of disparity and texture gradient information l.jpg
1 / 37
0
0
958 days ago, 316 views
PowerPoint PPT Presentation
Diagram. Foundation: Optimal sign combinationMethods: incline discriminationSingle-prompt resultsTwo-signal results: saw slantTwo-prompt results: JNDsConclusions. Layout. Foundation: Optimal prompt combinationMethods: incline discriminationSingle-signal resultsTwo-prompt results: saw slantTwo-prompt results: JNDsConclusions.

Presentation Transcript

Slide 1

Impacts of Viewing Geometry on Combination of Disparity and Texture Gradient Information James M. Hillis Michael S. Landy Martin S. Banks

Slide 2

Outline Background: Optimal sign mix Methods: incline segregation Single-prompt outcomes Two-prompt outcomes: saw incline Two-sign outcomes: JNDs Conclusions

Slide 3

Outline Background: Optimal prompt mix Methods: incline separation Single-sign outcomes Two-sign outcomes: saw incline Two-signal outcomes: JNDs Conclusions

Slide 4

Sources of Depth Information Motion Parallax Occlusion Stereo Disparity Shading Texture Linear Perspective Etc.

Slide 5

Depth Cues Motion Parallax Occlusion Stereo Disparity Shading Texture Linear Perspective Etc.

Slide 6

Optimal Cue Combination: Statistical Approach If the objective is to deliver a gauge with insignificant fluctuation, and the signs are uncorrelated, then the ideal gauge is a weighted normal where

Slide 7

Optimal Cue Combination: Bayesian Inference Approach From the Bayesian viewpoint, the estimations D and T each outcome in a probability capacity These are joined with an earlier dispersion

Slide 8

Optimal Cue Combination: Bayesian Inference Approach From Bayes govern, and expecting contingent freedom of the prompts, the back conveyance fulfills:

Slide 9

Optimal Cue Combination: Bayesian Inference Approach Finally, accepting Gaussian probabilities and earlier, things being what they are the most extreme a posteriori (MAP) assess fulfills: where p remains for the earlier which goes about as though it were an extra signal, and the weights are again corresponding to converse change.

Slide 10

Previous Qualitative Tests that Cue Weights Depend on Reliability Young, Landy & Maloney (1993) Johnston, Cumming & Landy (1994) Rogers and Bradshaw (1995) Frisby, Buckley & Horsman (1995) Backus and Banks (1999) and so forth and so forth.

Slide 11

Previous Quantitative Tests that Cue Weights Depend on Reliability Landy & Kojima (2001) – surface signals to area Ernst & Banks (2002) – visual and haptic prompts to estimate Gepshtein & Banks (2003) – visual and haptic signs to measure Knill & Saunders (2003) – surface and dissimilarity prompts to incline

Slide 12

The Current Study Texture and difference prompts to incline Vary unwavering quality by shifting base inclination (as in Knill & Saunders, 2003) and remove Measure single-signal dependability Compare two-prompt weights to expectations Compare two-signal dependability to forecasts

Slide 13

Outline Background: Optimal signal blend Methods: incline segregation Single-prompt outcomes Two-prompt outcomes: saw incline Two-prompt outcomes: JNDs Conclusions

Slide 14

Types of Stimuli Disparity-just: inadequate irregular specks Texture: Voronoi surfaces saw monocularly Two-signal jolts: Voronoi surface stereograms, both clash and no-contention

Slide 15

Stimuli – Disparity-just

Slide 16

Stimuli – Voronoi surfaces

Slide 17

Cue Conflict Stimuli

Slide 18

Methods Task: 2IFC inclination separation Single-sign and two-signal squares Opposite-sign inclinations blended crosswise over trials in a piece to evade incline adjustment One boost settled, other changed by staircase; a few interleaved staircases Analysis: fit psychometric capacity to assess PSE and JND

Slide 19

Outline Background: Optimal signal mix Methods: incline segregation Single-prompt outcomes Two-prompt outcomes: saw incline Two-signal outcomes: JNDs Conclusions

Slide 20

Single-signal JNDs: Texture

Slide 21

Single-sign JNDs: Disparity

Slide 22

Single-signal JNDs: Disparity

Slide 23

Predicted Cue Weights

Slide 24

Outline Background: Optimal sign mix Methods: incline segregation Single-signal outcomes Two-signal outcomes: saw incline Two-prompt outcomes: JNDs Conclusions

Slide 25

Cue Conflict Paradigm

Slide 26

Determination of PSEs

Slide 27

Determination of Weights

Slide 28

Full Two-Cue Dataset ACH JMH

Slide 29

Effect of Viewing Distance

Slide 30

Effect of Base Slant

Slide 31

Outline Background: Optimal sign mix Methods: incline segregation Single-sign outcomes Two-sign outcomes: saw incline Two-sign outcomes: JNDs Conclusions

Slide 32

Improvement in Reliability with Cue Combination If the ideal weights are utilized: then the subsequent fluctuation is lower than that accomplished by either prompt alone.

Slide 33

Improvement in JND with 2 Cues

Slide 34

Conclusion The information are reliable with ideal sign mix Texture weight is expanded with expanding separation and expanding base inclination, as anticipated Two prompt JNDs are for the most part lower than the constituent single-sign JNDs Thus, weights are resolved trial-by-trial, in view of the present jolt data and, specifically, the two single-prompt inclination gauges

Slide 36

Are Cue Weights Chosen Locally?

Slide 37

Are Cue Weights Chosen Locally?

SPONSORS