390 likes | 479 Views
An Interactive Background Blurring Mechanism and Its Applications. NTU CSIE Yan Chih -Yu Advisor: Wu Ja -Ling, Ph.D. Outline. Introduction Related Work Method Object Segmentation Depth Map Generation Image Defocus Experimental Result Applications Conclusion. Introduction.
E N D
An Interactive Background Blurring Mechanism and Its Applications NTU CSIE Yan Chih-Yu Advisor: Wu Ja-Ling, Ph.D.
Outline • Introduction • Related Work • Method • Object Segmentation • Depth Map Generation • Image Defocus • Experimental Result • Applications • Conclusion
Introduction • Depth of field • Circle of confusion Photo plane Real world
Introduction • Depth of field • Circle of confusion Photo plane Out of focus Real light Circle of Confusion
Introduction • Depth of field Focus plane Photo plane Out of focus – blur range Real light Depth of field Circle of Confusion - Readability range
Introduction • Shallow focus Photo plane Real world
Introduction • Shallow focus • Highlight the subject by softening background diffusion Shallow focus by DSLR NIKON D90(2008) Deep focus by DC NIKON E4300 (2003)
Depth Map Generation Object Segmentation Applications Image Defocus Perspective Box Lazy Snapping Camera Settings Pop-up Card Alpha Matting Defocus Blur Face Detection Privacy Preserving Photo Browser Method
Method User Stroke User Stroke Resize to 20% Resize to 20% • Segment • Lazy snapping • Mean shift • Alpha matting • Face detection Mean shift Mean shift Do hard graph cut by lazy snapping Do hard graph cut by lazy snapping Segmentation tri-Map Segmentation alpha Map Lazy Snapping. ACM Trans. On Graphics 2004. Mean Shift: A Robust Approach Toward Feature Space Analysis. PAMI 2002. A Bayesian Approach to Digital Matting. CVPR 2001.
Method Perspective Box Perspective Box Pop-up card Pop-up card • Depth map • Perspective Box • Vanish point • Rear wall Normal vector Normal vector Depth Map Depth Map Tour Into the Picture: Using a spidery mesh user interface to make animation from a single image. SIGGRAPH 1997
Method Perspective Box Pop-up card Pop-up card • Depth map • Pop-up card Normal vector Normal vector Depth Map Depth Map
Method Perspective Box Pop-up card • Depth map • Normal vector Normal vector Normal vector Depth Map Depth Map
(a) (b) ub vb ub vb b b d d u v u v Method Depth Map Segment Map Camera setting • Image defocus • Blur circle diameter Aperture size Focal length of the lens Distance of focus Blur circle diameter
Method • Defocus blur • bokeh
Method • Defocus blur
Method • Near-by object case Focus on the flower Focus on the flower
Method Blur circle diameter Objects in front of the focus Objects behind the focus • Near-by object case Blur shape as alpha map Text synthesis inpainting Defocus Blur Defocus Blur Interpolation by alpha map Shallow focus image
Experimental Result 1 • Defocus blur method proposed in the system comparing with other blur filter results Deep focus Shallow focus
Experimental Result 1 • Defocus blur method proposed in the system comparing with other blur filter results Gaussian blur Defocus blur Defocus blur + Bokeh
Experimental Result 1 Original photograph • Defocus blur method proposed in the system comparing with other blur filter results Gaussian blur Defocus blur Defocus blur + Bokeh
Experimental Result 2 • with / without depth variation in the background Real photograph taken by DSLR
Experimental Result 2 • with / without depth variation in the background Depth variation Original photograph Result after post-processing Depth fixed
Experimental Result 3 • Near-by object case Focus on the flower Focus on the flower
Experimental Result 3 • Near-by object case Interpolation result Without inpainting
Experimental Result 4 • Comparing with Photoshop Original image Photoshop (1 Hour) Our system (3~5min)
Applications - Image Transition at Photo Browser Movie with defocus blur effect
Applications - Image Transition at Photo Browser Concatenate two unrelated images
Conclusion • We proposed an interactive refocusing tool for background blurring • Simple user hint • Defocus blur kernel • Concatenate two related picture • Future work • concatenate two or more unrelated pictures • Color based image retrieval technique
Experimental Result 1 • Defocus blur method proposed in the system comparing with other blur filter results Deep focus Shallow focus Difference
Experimental Result 1 • Defocus blur method proposed in the system comparing with other blur filter results Defocus blur PSNR:22.79 Gaussian blur PSNR:20.34 Shallow focus Difference
END THANK YOU