Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
CSE 473/573 Computer Vision and Image Processing (CVIP) Ifeoma Nwogu [email protected] Lecture 19 – Dense motion estimation (OF) 1 Schedule • Last class – We finished stereo and multi-view geometry (high level) • Today – Optical flow • Readings for today: Forsyth and Ponce 10.6, 11.1.2 2 Visual motion Many slides adapted from S. Seitz, R. Szeliski, M. Pollefeys Motion and perceptual organization • Sometimes, motion is the only cue Motion and perceptual organization • Sometimes, motion is the only cue Motion and perceptual organization • Even “impoverished” motion data can evoke a strong percept G. Johansson, “Visual Perception of Biological Motion and a Model For Its Analysis", Perception and Psychophysics 14, 201-211, 1973. Motion and perceptual organization • Even “impoverished” motion data can evoke a strong percept G. Johansson, “Visual Perception of Biological Motion and a Model For Its Analysis", Perception and Psychophysics 14, 201-211, 1973. Motion and perceptual organization • Even “impoverished” motion data can evoke a strong percept YouTube video G. Johansson, “Visual Perception of Biological Motion and a Model For Its Analysis", Perception and Psychophysics 14, 201-211, 1973. Uses of motion • • • • Estimating 3D structure Segmenting objects based on motion cues Learning and tracking dynamical models Recognizing events and activities Classes of techniques for motion estimation • Feature-based methods – Extract visual features (corners, textured areas) and track them – Sparse motion fields, but possibly robust tracking – Suitable especially when image motion is large (10s of pixels) • Direct-methods – Directly recover image motion from spatio-temporal image brightness variations – Global motion parameters directly recovered without an intermediate feature motion calculation – Dense motion fields, but more sensitive to appearance variations – Suitable for video and when image motion is small (< 10 pixels) Szeliski Motion field • The motion field is the projection of the 3D scene motion into the image Patch based image motion How do we determine correspondences? I J Assume all change between frames is due to motion: J ( x, y ) I ( x u ( x, y ), y v( x, y )) Optical flow • Definition: optical flow is the apparent motion of brightness patterns in the image • Ideally, optical flow would be the same as the motion field • Have to be careful: apparent motion can be caused by lighting changes without any actual motion – Think of a uniform rotating sphere under fixed lighting vs. a stationary sphere under moving illumination Estimating optical flow I(x,y,t–1) I(x,y,t) • Given two subsequent frames, estimate the apparent motion field u(x,y) and v(x,y) between them • Key assumptions • Brightness constancy: projection of the same point looks the same in every frame • Small motion: points do not move very far • Spatial coherence: points move like their neighbors The brightness constancy constraint I(x,y,t–1) I(x,y,t) • Brightness Constancy Equation: I ( x, y, t 1) I ( x u ( x, y ), y v( x, y ), t ) Linearizing the right side using Taylor expansion: I ( x, y, t 1) I ( x, y, t ) I x u ( x, y ) I y v( x, y ) Hence, I x u I y v It 0 The brightness constancy constraint I x u I y v It 0 • How many equations and unknowns per pixel? – One equation, two unknowns • What does this constraint mean? I (u, v) I t 0 • The component of the flow perpendicular to the gradient (i.e., parallel to the edge) is unknown The brightness constancy constraint I x u I y v It 0 • How many equations and unknowns per pixel? – One equation, two unknowns • What does this constraint mean? I (u, v) I t 0 • The component of the flow perpendicular to the gradient (i.e., parallel to the edge) is unknown gradient (u,v) If (u, v) satisfies the equation, so does (u+u’, v+v’) if I (u ' , v' ) 0 (u’,v’) (u+u’,v+v’) edge The aperture problem Perceived motion The aperture problem Actual motion The barber pole illusion http://en.wikipedia.org/wiki/Barberpole_illusion The barber pole illusion http://en.wikipedia.org/wiki/Barberpole_illusion Solving the aperture problem • How to get more equations for a pixel? • Spatial coherence constraint: pretend the pixel’s neighbors have the same (u,v) – E.g., if we use a 5x5 window, that gives us 25 equations per pixel I (x i ) [u, v] I t (x i ) 0 I x (x1 ) I y (x1 ) I t (x1 ) I (x ) I (x ) I (x ) u y 2 t 2 x 2 v I x (x n ) I y (x n ) I t (x n ) B. Lucas and T. Kanade. An iterative image registration technique with an application to stereo vision. In Proceedings of the International Joint Conference on Artificial Intelligence, pp. 674–679, 1981. Solving the aperture problem • Least squares problem: I x (x1 ) I y (x1 ) I t (x1 ) I (x ) I (x ) I ( x ) y 2 u t 2 x 2 v I x (x n ) I y (x n ) I t (x n ) • When is this system solvable? • What if the window contains just a single straight edge? B. Lucas and T. Kanade. An iterative image registration technique with an application to stereo vision. In Proceedings of the International Joint Conference on Artificial Intelligence, pp. 674–679, 1981. Conditions for solvability • “Bad” case: single straight edge Conditions for solvability • “Good” case Lucas-Kanade flow • Linear least squares problem I x (x1 ) I y (x1 ) I t (x1 ) I (x ) I (x ) I (x ) u y 2 x 2 t 2 v I ( x ) I ( x ) x n y n I t (x n ) AUd n2 21 n1 • Solution given by ( A T A)U A T d I x I x I x I y I I I x I t I u I y I t yIy v x y • The summations are over all pixels in the window B. Lucas and T. Kanade. An iterative image registration technique with an application to stereo vision. In Proceedings of the International Joint Conference on Artificial Intelligence, pp. 674–679, 1981. Lucas-Kanade flow I x2 I x I y I I I x y 2 y u I x It I I v y t • Recall the Harris corner detector: M = ATA is the second moment matrix • We can figure out whether the system is solvable by looking at the eigenvalues of the second moment matrix • The eigenvectors and eigenvalues of M relate to edge direction and magnitude • The eigenvector associated with the larger eigenvalue points in the direction of fastest intensity change, and the other eigenvector is orthogonal to it Visualization of second moment matrices Visualization of second moment matrices Interpreting the eigenvalues Classification of image points using eigenvalues of the second moment matrix: 2 1 and 2 are small “Edge” 2 >> 1 “Flat” region “Corner” 1 and 2 are large, 1 ~ 2 “Edge” 1 >> 2 1 Visualization of second moment matrices The Aperture Problem Let M I I T and I x I t b I y I t • Algorithm: At each pixel compute U by solving MU b • M is singular if all gradient vectors point in the same direction • e.g., along an edge • of course, trivially singular if the summation is over a single pixel or there is no texture • i.e., only normal flow is available (aperture problem) • Corners and textured areas are OK Szeliski Example * From Khurram Hassan-Shafique CAP5415 Computer Vision 2003 Uniform region – gradients have small magnitude – small 1, small 2 – system is ill-conditioned SSD – uniform region Edge – gradients have one dominant direction – large 1, small 2 – system is ill-conditioned SSD Surface -- edge High-texture or corner region – gradients have different directions, large magnitudes – large 1, large 2 – system is well-conditioned SSD Surface – textured area or corner Optical Flow Results * From Khurram Hassan-Shafique CAP5415 Computer Vision 2003 Errors in Lucas-Kanade • Fails when intensity structure in window is poor • The motion is large (larger than a pixel) – Iterative refinement – Coarse-to-fine estimation – Exhaustive neighborhood search (feature matching) • A point does not move like its neighbors – Motion segmentation • Brightness constancy does not hold – Exhaustive neighborhood search with normalized correlation Coarse-to-Fine Estimation warp + a aw J pixels refine u=1.25 u=2.5 pixels Δa u=5 pixels image J Pyramid of image J u=10 pixels image I Pyramid of image I Szeliski Coarse-to-Fine Estimation ain J J warp + pyramid construction J I Jw a a warp Jw a warp + pyramid construction I refine + J I refine aout Jw refine I a Szeliski Multi-resolution registration * From Khurram Hassan-Shafique CAP5415 Computer Vision 2003 Optical Flow Results * From Khurram Hassan-Shafique CAP5415 Computer Vision 2003 Optical Flow Results * From Khurram Hassan-Shafique CAP5415 Computer Vision 2003 State-of-the-art optical flow Start with something similar to Lucas-Kanade + gradient constancy + energy minimization with smoothing term + region matching + keypoint matching (long-range) Region-based +Pixel-based +Keypoint-based Large displacement optical flow, Brox et al., CVPR 2009 Source: J. Hays Feature tracking • So far, we have only considered optical flow estimation in a pair of images • If we have more than two images, we can compute the optical flow from each frame to the next • Given a point in the first image, we can in principle reconstruct its path by simply “following the arrows” Tracking challenges • Ambiguity of optical flow – Need to find good features to track • Large motions, changes in appearance, occlusions, disocclusions – Need mechanism for deleting, adding new features • Drift – errors may accumulate over time – Need to know when to terminate a track Shi-Tomasi feature tracker • Find good features using eigenvalues of secondmoment matrix – Key idea: “good” features to track are the ones whose motion can be estimated reliably • From frame to frame, track with Lucas-Kanade – This amounts to assuming a translation model for frame-toframe feature movement • Check consistency of tracks by affine registration to the first observed instance of the feature – Affine model is more accurate for larger displacements – Comparing to the first frame helps to minimize drift J. Shi and C. Tomasi. Good Features to Track. CVPR 1994. Tracking example J. Shi and C. Tomasi. Good Features to Track. CVPR 1994. Non Gaussian noise • Least square solution assumes error in the image motion estimation are Gaussian in nature • The matrix M or the structured tensor matrix is computed using finite difference methods – forward, backward, and central differences – Can obtain higher order evaluations based on how the derivatives are computed (e.g adaptive windowing etc.) 52 Robust Estimation Noise distributions are often non-Gaussian, having much heavier tails. Noise samples from the tails are called outliers. • Sources of outliers (multiple motions): – specularities / highlights – jpeg artifacts / interlacing / motion blur – multiple motions (occlusion boundaries, transparency) u2 velocity space + + u1 Black Occlusion occlusion disocclusion shear Multiple motions within a finite region. Black Coherent Motion Possibly Gaussian. Black Multiple Motions Definitely not Gaussian. Black Layered Scene Representations Motion representations • How can we describe this scene? Szeliski Block-based motion prediction • Break image up into square blocks • Estimate translation for each block • Use this to predict next frame, code difference (MPEG-2) Szeliski Layered motion • Break image sequence up into “layers”: • = • Describe each layer’s motion Szeliski Layered motion • Advantages: • can better handle occlusions / disocclusions • each layer’s motion can be smooth • can be used for video segmentation in semantic processing • Difficulties: • how to determine the correct number of layers? • how to assign pixels? • how to model the layer motion? Szeliski Layers for video summarization Szeliski Background modeling (MPEG-4) • Convert masked images into a background sprite for layered video coding • + + + • = Szeliski What are layers? • [Wang & Adelson, 1994; Darrell & Pentland 1991] • intensities • alphas • velocities Szeliski Fragmented Occlusion Results Results How to estimate the layers 1. 2. 3. 4. 5. compute coarse-to-fine flow estimate affine motion in blocks (regression) cluster with k-means assign pixels to best fitting affine region re-estimate affine motions in each region… Szeliski Layer synthesis • For each layer: • • • stabilize the sequence with the affine motion compute median value at each pixel Determine occlusion relationships Szeliski Results Szeliski Recent GPU Implementation • http://gpu4vision.icg.tugraz.at/ • Real time flow exploiting robust norm + regularized mapping Recent results: SIFT Flow Slide Credits • Svetlana Lazebnik – UIUC • Trevor Derrell – UC Berkeley 73 Next class • Segmentation via clustering • Readings for next lecture: – Forsyth and Ponce chapter 9 – Szelinski chapter 5 • Readings for today: – Forsyth and Ponce 10.6 and 11.1.2 74 Questions 75