Combining Laser Scans Yong Joo Kil 1, Boris Mederos 2, and Nina Amenta 1 1 Department of Computer Science, University of California at Davis 2 Instituto Nacional de Matematica Pura e Aplicada - IMPA IDAV Institute for Data Analysis and Visualization Visualization and Graphics Research Group
2D Super Resolution A Fast Super-Resolution Reconstruction Algorithm, [Michael Elad, Yacov Hel-Or] Low Resolution ImagesSuper Resolution Image
Surface Super Resolution One Raw Scan Super resolved (100 scans) Photo
Improve 3D Acquisition Methods Better hardware –Costly Multiple scans + software –Refine output of current hardware –Cost effective –Smaller devices
Physical Setup xyxy z (viewing direction) Minolta Vivid 910
3D Super Resolution Pipeline Input Scans Global Registration Super Resolution Super Registration Convergence No Yes Smoothing Super Resolution Mesh
Viewing direction axis z x y
Sample Points Low Resolution Sample Spacing Width Of one Scan
Super Resolution Sample Spacing q N(q) width/4
2.5D Super Resolution
First Super Resolution Mesh (S 1 )
Super Resolution Method Input Scans Global Registration Super Resolution Super Registration Convergence No Yes Smoothing Super Resolution Mesh
Bilateral Filter
Super Resolution Method Input Scans Global Registration Super Resolution Super Registration Convergence No Yes Smoothing Super Resolution Mesh
Super Registration raw scan super resolution mesh
Second Super Resolution Mesh S 2
Super Resolution Method Input Scans Global Registration Super Resolution Super Registration Convergence No Yes Smoothing Super Resolution Mesh
Point Samples (1 st Model) Derived from Super-Resolution Reconstruction of Images - Static and Dynamic Paradigms [Michael Elad] Nyquist Sampling Theorem: Sample signal finely enough, then Reconstruct original signal perfectly. Band limited signal
Sampling at lower resolution Derived from Super-Resolution Reconstruction of Images - Static and Dynamic Paradigms [Michael Elad] That’s it!
Linear Model with Blur (2 nd Model) High- Resolution Image X Derived from Super-Resolution Reconstruction of Images - Static and Dynamic Paradigms [Michael Elad] C Blur 1 D 1 Decimation Low- Resolution Images Transformation F 1 Y 1 E 1 Noise + C N F N D N Y N E N +
The Model as One Equation Derived from Super-Resolution Reconstruction of Images - Static and Dynamic Paradigms [Michael Elad]
Model for 3D laser scan?
Pipeline : Laser Scanner Derived from Better Optical Triangulation through Spacetime Analysis, Curless and Levoy, 1995 laser beam Surface Peak reconstruction CCD sensor
Video sequence x y time
Non Linear functions
Simplification Assume –Points from Surface –Gaussian Noise
Point Sampling Model High- Resolution Image X C Blur k D k Decimation Low- Resolution Images Transformation F k x [ ELAD M., HEL-OR Y.: A fast super-resolution reconstruction algorithm for pure translational motion and common space invariant blur. IEEE Transactions on Image Processing 10,8 (2001) ] Solution Average Y k E k Gaussian Noise +
Simplification Solution –Register scans –Averaging Easy Inexpensive It works!
Close-up Scan of Parrot 146 Scans 4 times the original resolution.
Super resolve far & close objects? Derived from Better Optical Triangulation through Spacetime Analysis, Curless and Levoy, 1995 Surface CCD sensor
Super resolve small & large objects? One raw ScanSuper resolution (117 scans)
Is it worth taking more than one scan? One raw scan Super resolution PhotographSubdivion of (a)
Is it worth shifting? With Shifts (117scans)Without Shifts (117scans)
How many scans are enough?
Point Distribution
Tiling Artifact
Sampling Pattern Random xy shift + Rotation
Mayan Tablet (One Scan)
39 Mayan Tablet (90 scans)
40 Before & After
41 Systematic Errors Super resolvedPhoto
42 Parrot Model (6 views * 100 scans)
Future work 2.5D to 3D Resolving Systematic Errors Other Devices
Acknowledgements Kelcey Chen Geomagic Studios NSF CCF Brazilian National Council of Technological and Scientific Development (CNPq)
45 Extras
Interpolations
Nyquist frequency
48 Data
50 Solving this linear system is equivalent to an average. [ ELAD M., HEL-OR Y.: A fast super-resolution reconstruction algorithm for pure translational motion and common space invariant blur. IEEE Transactions on Image Processing 10,8 (2001) ] Mimize Diagonal Matrix Can be a permutation or displacement matrix Equivalent to
51 Error between low res and super res.
52 Error between low res and super res.
53 Registeration result
54 Before and After Registration
55 Error between low res and super res.
56 Least Squares Minimize: Solve by:, or Steepest Descent Iteration:,