Local Laplacian Filters: Edge-aware Image Processing with a Laplacian Pyramid Paper by Sylvain Paris, Samuel W. Hasinoff, Jan Kautz Presenter: Jing Niu
An Example ● Input: Milestones and Advances in Image Analysis WS 12/13 2
An Example ● output Milestones and Advances in Image Analysis WS 12/13 3
Outline ● Motivation ● Laplacian Pyramids ● Local Laplacian Filtering ● Algorithm ● Applications Milestones and Advances in Image Analysis WS 12/13 4
Motivation Belived to be unsuitable for: ● Representing edges ● Edge-aware operations (edge-preserving smoothing, tone ● mapping) Reason: ● – Build upon isotropic, spatially invariant gaussian kernel Goal: ● Flexible approach ● edge-aware image processing using ● – simple point-wise manipulation of Laplacian pyramids Milestones and Advances in Image Analysis WS 12/13 5
Laplacian and Guassian Pyramids ● Gaussian Pyramid: ● A set of image levels ● Represent lower resolution upsample subsample ● High frequency details disappear Milestones and Advances in Image Analysis WS 12/13 6
Laplacian Pyramid ● Downsampling:decomposition G 0 G 1 G 2 Residual L 1 Ref[1] L 0 Milestones and Advances in Image Analysis WS 12/13 7
Laplacian Pyramid ● Upsampling: G 0 G 1 G 2 L 1 L 0 Ref[1] Milestones and Advances in Image Analysis WS 12/13 8
Local Laplacian Filtering ● Range compression and clipping Input Signal Milestones and Advances in Image Analysis WS 12/13 9
Local Laplacian Filtering ● Range compression and clipping Input Signal Right clippling Milestones and Advances in Image Analysis WS 12/13 10
Local Laplacian Filtering ● Range compression and clipping Input Signal Right clippling Milestones and Advances in Image Analysis WS 12/13 11
Local Laplacian Filtering ● Range compression and clipping Right clipping Input Signal Left Clipping Milestones and Advances in Image Analysis WS 12/13 12
Local Laplacian Filtering ● Range compression and clipping Input Signal Right clipping Left clipping merged Milestones and Advances in Image Analysis WS 12/13 13
Point-wise Remapping function edge--aware tone manipulation edge--aware detail manipulation tone mapping inverse tone mapping detail smoothing detail enhancement combined operator detail enhance + tone map Milestones and Advances in Image Analysis WS 12/13 14
An overview of the algorithm Approach: construct laplacian pyramid of filtered output Milestones and Advances in Image Analysis WS 12/13 15
Illustration Milestones and Advances in Image Analysis WS 12/13 16
Illustration Milestones and Advances in Image Analysis WS 12/13 17
Illustration Milestones and Advances in Image Analysis WS 12/13 18
Illustration Milestones and Advances in Image Analysis WS 12/13 19
Illustration Milestones and Advances in Image Analysis WS 12/13 20
Illustration Milestones and Advances in Image Analysis WS 12/13 21
Illustration Milestones and Advances in Image Analysis WS 12/13 22
Application ● Detail manipulation ● Tone mapping Milestones and Advances in Image Analysis WS 12/13 23
Application Detail manipulation ● Tone mapping ● β, σ r similar effects on tone mapping results α is set to 1 Milestones and Advances in Image Analysis WS 12/13 24
More Results bilateral filter and close up Our result and close up Milestones and Advances in Image Analysis WS 12/13 25
More Results Milestones and Advances in Image Analysis WS 12/13 26
Conclusion ● Edge aware ● Based solely on laplacian pyramid ● Simple method ● Robustness ● Artifact-free ● high quality image ● open new perspectives on multi-scale image analysis and editing Milestones and Advances in Image Analysis WS 12/13 27
Reference ● Pyramid-based Image Synthesis Theory Shuguang Mao and Morgan Brown ● Advanced Image Analysis Christian Schmaltz ● Local Laplacian Filters: Edge-aware Image Processing with a Laplacian Pyramid Sylvain Paris, Samuel W. Hasinoff, Jan Kautz Milestones and Advances in Image Analysis WS 12/13 28
Thank you Milestones and Advances in Image Analysis WS 12/13 29
Recommend
More recommend
