|
Orientation fields and 3D shape estimation
|
tarix | 30.10.2018 | ölçüsü | 33,12 Mb. | | #76228 |
|
Roland W. Fleming Max Planck Institute for Biological Cybernetics
Approach I: inverse optics
Real-world illumination is highly structured Real-world illumination is highly structured Specular reflections of the real world are a bit like texture Can we solve the 3D shape of mirrors using shape-from-texture ?
Slant distorts texture but not reflections
Image distortions
Image distortions
Image distortions Curvature distorts reflections but not texture
Curvatures determine distortions
Curvatures determine distortions
Shape-from-texture and shape-from-specularity follow different rules - For texture, image compression depends on surface slant
- first derivative of surface
- For reflections, image compression depends on surface curvature properties
Local analysis: banding patterns
Gauge Figure Task Subject adjusts 3D orientation of “gauge figure” to match local orientation of surface
Slant and Tilt
Results I
Results II
Is it just the occluding contour?
Interpreting distorted reflections
Population codes
Population codes
Population codes
Population codes
Orientation fields
Orientation fields are robust
Beyond specularity
Differences between diffuse and specular reflection
Beyond specularity
Latent orientation structure
Orientation fields in shading
Reflectance as Illumination
Texture
Texture
Orientation fields in texture
Affine Transformation
Affine Transformation
Affine Transformation
Affine Transformation
Illusory distortions of shape
Potential of Orientation Fields Uses biologically plausible measurements
Potential of Orientation Fields No need for visual system to estimate reflectance or illumination explicitly.
Potential of Orientation Fields Stable across albedo discontinuities.
Potential of Orientation Fields Handle improbable combinations of reflectance and illumination.
Potential of Orientation Fields May explain how images with no obvious BRDF interpretation nevertheless yield 3D percepts
Converting between cues
Matte vs. Shiny Same generative statistics, different mappings
Texture vs. Reflectance
Texture vs. Reflectance
Texture vs. Reflectance
Texture vs. Reflectance
Conclusions Orientation fields are potentially a very powerful source of information about 3D shape For the early stages of 3D shape processing, seemingly different cues may have more in common than previously thought
Todd’s Blobs
Todd’s Blobs
What still needs to be explained? For Lambertian materials (or blurry illuminations), the reflectance map is so smooth that it is significantly anisotropic. Therefore shading orientation fields vary considerably with changes in illumination.
What still needs to be explained?
Two possibilities Change in orientation field predicts (subtle) changes in perceived 3D shape There are higher-order invariants in the orientation fields
Synthesis Koenderink, J. J., & van Doorn, A. J. (1980). Photometric invariants related to solid shape. Optica Acta, 27(7), 981-996. Li, A., & Zaidi, Q. (2000). Perception of three-dimensional shape from texture is based on patterns of oriented energy. Vision Research, 40(2), 217-242. Malik, J., & Rosenholtz, R. (1997). Computing local surface orientation and shape from texture for curved surfaces. International Journal of Computer Vision, 23, 149-168. Breton, P., & Zucker, S. W. (1996). Shadows and shading flow fields. In Proceedings of CVPR, 782-789.[[AUTHOR: Please provide the publisher and location for this book.]] San Francisco, CA. Ben-Shahar, O., & Zucker, S. (2001). On the perceptual organization of texture and shading flows: From a geometrical model to coherence computation. In Proceedings of CVPR, pp. 1048-1055.[[AUTHOR: Is this a journal? If so, please provide the volume. If it’s a meeting, please provide the city, state, where the meeting occurred. Thanks.]] Kawaii, HI. Huggins, P. S., Chen, H. F., Belhumeur, P. N., & Zucker, S. W. (2001). Finding folds: On the Appearance and Identification of Occlusion, in CVPR'01. Proceedings IEEE Conference on Computer Vision and Pattern Recognition, 2, 718-725. Grossberg, S. & Mingolla, E. (1987). Neural Dynamics of Surface Perception:Boundary Webs , Illuminants, and Shape from Shading. Commun. Vis. Graph.Image Proc., 37:116—165.
Part II: Conclusions We can accurately and reliably perceive the 3D shape of perfect mirrors - Across illuminations
- In the absence of context to specify the illumination
- Subjects aren’t just using the occluding boundary to perform the task
We think subjects estimate shape directly from the pattern of distortions across the image. The pattern of distortions: - can be extracted from the image by populations of simple oriented filters (cf. V1 cell receptive fields)
- are diagnostic of shape
- remain quite stable across changes in the illumination
Plastics
When the world is anisotropic
Stripy world
Counter-example I: blurring Blurred versions have sparser edge-maps but still look solid
Counter-example II: negatives Photographic negatives have identical edge-maps but look somewhat flatter
Matte vs. Shiny Same generative statistics, different mappings
Orientation maps
Effects of compression
Shape from texture
Illusory distortions of shape
Dostları ilə paylaş: |
|
|