Unformatted text preview:

C280, Computer VisionProf. Trevor [email protected] 7: TextureLast Time: Feature Detection and Matching• Local features• Pyramids for invariant feature detection•Invariant descriptors•Invariant descriptors• MatchingToday: TextureWhat defines a texture?Includes: more regular patternsIncludes: more random patternsScale: objects vs. textureOften the same thing in the world can occur as texture or an object, depending on the scale we are considering.Why analyze texture?Importance to perception:• Often indicative of a material’s properties• Can be important appearance cue, especially if shape is similar across objects•Aim to distinguish between shape, boundaries, and •Aim to distinguish between shape, boundaries, and textureTechnically: • Representation-wise, we want a feature one step above “building blocks” of filters, edges.Texture-related tasks• Shape from texture– Estimate surface orientation or shape from image textureShape from texture• Use deformation of texture from point to point to estimate surface shapePics from A. Loh: http://www.csse.uwa.edu.au/~angie/phdpics1.htmlTexture-related tasks• Shape from texture– Estimate surface orientation or shape from image texture• Segmentation/classification from texture cues–Analyze, represent texture–Analyze, represent texture– Group image regions with consistent texture• Synthesis– Generate new texture patches/images given some exampleshttp://animals.nationalgeographic.com/Color vs. textureRecall: These looked very similar in terms of their color distributions (when our features were R-G-B)But how would their texture distributions compare?Psychophysics of texture• Some textures distinguishable with preattentive perception–without scrutiny, eye movements [Julesz 1975]Same or different?Julesz• Textons: analyze the texture in terms of statistical relationships between fundamental texture elements, called “textons”. •It generally required a human to look at the •It generally required a human to look at the texture in order to decide what those fundamental units were...Texture representation• Textures are made up of repeated local patterns, so:– Find the patterns• Use filters that look like patterns (spots, bars, raw patches…)•Consider magnitude of response•Consider magnitude of response– Describe their statistics within each local window• Mean, standard deviation• Histogram• Histogram of “prototypical” feature occurrencesTexture representation: examplemean d/dxvalue mean d/dyvalue Win. #1 4 10original imagederivative filter responses, squaredstatistics to summarize patterns in small windows …Texture representation: examplemean d/dxvalue mean d/dyvalue Win. #1 4 10Win.#2187original imagederivative filter responses, squaredstatistics to summarize patterns in small windows Win.#2187…Texture representation: examplemean d/dxvalue mean d/dyvalue Win. #1 4 10Win.#2187original imagederivative filter responses, squaredstatistics to summarize patterns in small windows Win.#2187…Texture representation: examplemean d/dxvalue mean d/dyvalue Win. #1 4 10Win.#2187original imagederivative filter responses, squaredstatistics to summarize patterns in small windows Win.#2187Win.#9 20 20……Texture representation: examplemean d/dxvalue mean d/dyvalue Win. #1 4 10Win.#2187Dimension 2 (mean d/dy value)statistics to summarize patterns in small windows Win.#2187Win.#9 20 20……Dimension 1 (mean d/dx value)Dimension 2 (mean d/dy value)Texture representation: examplemean d/dxvalue mean d/dyvalue Win. #1 4 10Win.#2187Dimension 2 (mean d/dy value)Windows with primarily horizontal edgesBothstatistics to summarize patterns in small windows Win.#2187Win.#9 20 20……Dimension 1 (mean d/dx value)Dimension 2 (mean d/dy value)Windows with small gradient in both directionsWindows with primarily vertical edgesTexture representation: exampleoriginal imagederivative filter responses, squaredvisualization of the assignment to texture “types”Texture representation: examplemean d/dxvalue mean d/dyvalue Win. #1 4 10Win.#2187Dimension 2 (mean d/dy value)Far: dissimilar texturesstatistics to summarize patterns in small windows Win.#2187Win.#9 20 20……Dimension 1 (mean d/dx value)Dimension 2 (mean d/dy value)Close: similar texturesTexture representation: window scale• We’re assuming we know the relevant window size for which we collect these statistics.Possible to perform scale Possible to perform scale selection by looking for window scale where texture description not changing.Filter banks• Our previous example used two filters, and resulted in a 2-dimensional feature vector to describe texture in a window.– x and y derivatives revealed something about local structure.structure.• We can generalize to apply a collection of multiple (d) filters: a “filter bank”• Then our feature vectors will be d-dimensional.– still can think of nearness, farness in feature spaced-dimensional features. . .. . .2d3dFilter banksscalesorientations• What filters to put in the bank?– Typically we want a combination of scales and orientations, different types of patterns.Matlab code available for these examples: http://www.robots.ox.ac.uk/~vgg/research/texclass/filters.htmlInfluential paper:Bergen and Adelson, Nature 1988Learn: use filters.Malik and PeronaLearn: use lots of filters, multi-ori&scale.Malik J, Perona P. Preattentive texture discrimination with early vision mechanisms. J OPT SOC AM A 7: (5) 923-932 MAY 1990If matching the averaged squared filter values is a good way to match a given texture, then maybe matching the entire marginal distribution (eg, the histogram) of a filter’s distribution (eg, the histogram) of a filter’s response would be even better.Jim Bergen proposed this…SIGGRAPH 1994Histogram matching algorithm“At this im1 pixel value, 10% of the im1 values are lower. What im2 pixel value has 10% of the im2 values below it?”Heeger-Bergen texture synthesis algorithmAlternate matching the histograms of all the subbands and matching the histograms of the reconstructed images.Bergen and HeegerLearn: use filter marginal statistics.Bergen and Heeger resultsBergen and Heeger failuresDe Bonet (and Viola)SIGGRAPH 1997DeBonetLearn: use filter conditional statistics across scale.DeBonetDeBonetWhat we’ve learned from the previous texture synthesis methodsFrom Adelson and Bergen:examine filter outputsFrom


View Full Document

Berkeley COMPSCI C280 - Lecture 7 - Texture

Download Lecture 7 - Texture
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Lecture 7 - Texture and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Lecture 7 - Texture 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?