Three-Dimensional Video Postproduction and Processing Ibraheem Alhashim - July 10 th 2013 CMPT 880.

Slides:



Advertisements
Similar presentations
Free-viewpoint Immersive Networked Experience February 2010.
Advertisements

Exploration of advanced lighting and shading techniques
Seeing 3D from 2D Images. How to make a 2D image appear as 3D! ► Output and input is typically 2D Images ► Yet we want to show a 3D world! ► How can we.
Anahita: A System for 3D Video Streaming with Depth Customization
Light-field 3DTV Research Péter Tamás Kovács Holografika.
December 5, 2013Computer Vision Lecture 20: Hidden Markov Models/Depth 1 Stereo Vision Due to the limited resolution of images, increasing the baseline.
3D Displays Duncan Lindbo, Rebecca Brown, Bao Khang Nguyen.
Boundary matting for view synthesis Samuel W. Hasinoff Sing Bing Kang Richard Szeliski Computer Vision and Image Understanding 103 (2006) 22–32.
A Novel 2D-to-3D Conversion System Using Edge Information IEEE Transactions on Consumer Electronics 2010 Chao-Chung Cheng Chung-Te li Liang-Gee Chen.
Advanced Computer Vision Introduction Goal and objectives To introduce the fundamental problems of computer vision. To introduce the main concepts and.
Copyright  Philipp Slusallek Cs fall IBR: Model-based Methods Philipp Slusallek.
MUltimo3-D: a Testbed for Multimodel 3-D PC Presenter: Yi Shi & Saul Rodriguez March 14, 2008.
Real-Time Geometric and Color Calibration for Multi-Projector Displays Christopher Larson, Aditi Majumder Large-Area High Resolution Displays Motivation.
Multi-view stereo Many slides adapted from S. Seitz.
High-Quality Video View Interpolation
Stereo and Multiview Sequence Processing. Outline Stereopsis Stereo Imaging Principle Disparity Estimation Intermediate View Synthesis Stereo Sequence.
1 Lecture 11 Scene Modeling. 2 Multiple Orthographic Views The easiest way is to project the scene with parallel orthographic projections. Fast rendering.
The Graphics Pipeline CS2150 Anthony Jones. Introduction What is this lecture about? – The graphics pipeline as a whole – With examples from the video.
A Novel 2D To 3D Image Technique Based On Object- Oriented Conversion.
1 Angel: Interactive Computer Graphics 4E © Addison-Wesley 2005 Models and Architectures Ed Angel Professor of Computer Science, Electrical and Computer.
2010/10/13VCLAB, National Tsing Hua University, Taiwan1 Multiview Video Kai-Chao Yang.
Project 4 Results Representation – SIFT and HoG are popular and successful. Data – Hugely varying results from hard mining. Learning – Non-linear classifier.
Research & Innovation 1 An Industry Perspective on VVG Research Oliver Grau BBC Research & Innovation VVG SUMMER SCHOOL '07.
Computer Vision Spring ,-685 Instructor: S. Narasimhan WH 5409 T-R 10:30am – 11:50am Lecture #15.
3D Stereo Video Coding Heejune AHN Embedded Communications Laboratory Seoul National Univ. of Technology Fall 2013 Last updated
3D/Multview Video. Outline Introduction 3D Perception and HVS 3D Displays 3D Video Representation Compression.
CAP4730: Computational Structures in Computer Graphics 3D Concepts.
Stereoscopic Analyzer On-Set Assistance System for 3D Capturing Frederik Zilly.
ICPR/WDIA-2012 High Quality Novel View Synthesis Based on Low Resolution Depth Image and High Resolution Color Image Jui-Chiu Chiang, Zheng-Feng Liu, and.
THE UNIVERSITY OF BRITISH COLUMBIA Random Forests-Based 2D-to- 3D Video Conversion Presenter: Mahsa Pourazad M. Pourazad, P. Nasiopoulos, and A. Bashashati.
Computer Graphics An Introduction. What’s this course all about? 06/10/2015 Lecture 1 2 We will cover… Graphics programming and algorithms Graphics data.
09/09/03CS679 - Fall Copyright Univ. of Wisconsin Last Time Event management Lag Group assignment has happened, like it or not.
Presenter: Pia Maffei Autostereoscopy and Film Pre-Viz and Promotion.
Advanced Computer Technology II FTV and 3DV KyungHee Univ. Master Course Kim Kyung Yong 10/10/2015.
© 2011 Adobe Systems Incorporated. All Rights Reserved. Adobe Confidential. Hands-on Introduction to After Effects Chris Jackson Author, Designer, Professor.
High-Resolution Interactive Panoramas with MPEG-4 발표자 : 김영백 임베디드시스템연구실.
CSC 461: Lecture 3 1 CSC461 Lecture 3: Models and Architectures  Objectives –Learn the basic design of a graphics system –Introduce pipeline architecture.
December 4, 2014Computer Vision Lecture 22: Depth 1 Stereo Vision Comparing the similar triangles PMC l and p l LC l, we get: Similarly, for PNC r and.
Computer Vision Why study Computer Vision? Images and movies are everywhere Fast-growing collection of useful applications –building representations.
1 Introduction to Computer Graphics with WebGL Ed Angel Professor Emeritus of Computer Science Founding Director, Arts, Research, Technology and Science.
1 Perception, Illusion and VR HNRS 299, Spring 2008 Lecture 8 Seeing Depth.
1Computer Graphics Lecture 4 - Models and Architectures John Shearer Culture Lab – space 2
Computer Vision Michael Isard and Dimitris Metaxas.
Stereo Viewing Mel Slater Virtual Environments
1 Artificial Intelligence: Vision Stages of analysis Low level vision Surfaces and distance Object Matching.
Computer Vision Lecture #10 Hossam Abdelmunim 1 & Aly A. Farag 2 1 Computer & Systems Engineering Department, Ain Shams University, Cairo, Egypt 2 Electerical.
CSE 185 Introduction to Computer Vision Stereo. Taken at the same time or sequential in time stereo vision structure from motion optical flow Multiple.
The Limitations of Stereoscopic 3D Display Technologies as a Path to Immersive Reality Aris Silzars, PhD Northlight Displays.
Taking stereoscopic tours of astronomical scenes Stuart Levy with Robert Patterson Advanced Visualization Lab - AVL at NCSA Nat'l Center for Supercomputing.
Subject Name: Computer Graphics Subject Code: Textbook: “Computer Graphics”, C Version By Hearn and Baker Credits: 6 1.
Immersive Rendering. General Idea ► Head pose determines eye position  Why not track the eyes? ► Eye position determines perspective point ► Eye properties.
Graphics II “3D” Graphics Cameron Miller INFO410 & INFO350 S INFORMATION SCIENCE Visual Computing.
DIGITAL CONTENT CREATION PROCESS fff PRE-PRODUCTION PRODUCTION POST-PRODUCTION Process ANIMATION PROCESS.
Perception and VR MONT 104S, Fall 2008 Lecture 8 Seeing Depth
Image-Based Rendering Geometry and light interaction may be difficult and expensive to model –Think of how hard radiosity is –Imagine the complexity of.
1 E. Angel and D. Shreiner: Interactive Computer Graphics 6E © Addison-Wesley 2012 Models and Architectures 靜宜大學 資訊工程系 蔡奇偉 副教授 2012.
An H.264-based Scheme for 2D to 3D Video Conversion Mahsa T. Pourazad Panos Nasiopoulos Rabab K. Ward IEEE Transactions on Consumer Electronics 2009.
AUDIO VIDEO SYSTEMS Prepared By :- KISHAN DOSHI ( ) PARAS BHRAMBHATT ( ) VAIBHAV SINGH THAKURALE ( )
1 2D TO 3D IMAGE AND VIDEO CONVERSION. INTRODUCTION The goal is to take already existing 2D content, and artificially produce the left and right views.
Digital Video Representation Subject : Audio And Video Systems Name : Makwana Gaurav Er no.: : Class : Electronics & Communication.
Visual Information Processing. Human Perception V.S. Machine Perception  Human perception: pictorial information improvement for human interpretation.
Processing visual information for Computer Vision
Padmasri Dr.BV Raju Institute Of Technology
A Novel 2D-to-3D Conversion System Using Edge Information
Jun Shimamura, Naokazu Yokoya, Haruo Takemura and Kazumasa Yamazawa
3D TV TECHNOLOGY.
Models and Architectures
Models and Architectures
Common Classification Tasks
Coding Approaches for End-to-End 3D TV Systems
Presentation transcript:

Three-Dimensional Video Postproduction and Processing Ibraheem Alhashim - July 10 th 2013 CMPT 880

Overview 1  History + Fundamentals  3D display technologies  Basic processing  View synthesis  2D to 3D conversion  3D media for mobile devices  Outlook

Overview 2  History + Fundamentals  3D display technologies  Basic processing  View synthesis  2D to 3D conversion  3D media for mobile devices  Outlook

A bit of history 3  Imaging technology James Maxwell 1855 JC d'Almeida 1858 Joseph Niépce 1826 Stereoscope 1860 Underwood 1901 The Power of Love (1922 film) The Jazz Singer 1927

A bit of history 4 Becky Sharp (1935)

A bit of history 5  Timeline of 3D movies 1950 s 1980 s 2010 s Worldwide: $2,782,275,172

3D Feature Films 6  Highest-grossing films 2012: nine of the top 15 were in 3D [businessinsider.com]  Industry forecast > 20% of TVs by 2015 Year3D Films

Other uses 7  WW2 maps  Virtual reality  3D Electron Microscopy  Video games + virtual cinema

3D Video 8  Usually marketed as objects popping off screen

3D Video 9  In reality.. It’s the same old concept Present slightly different image per eye The brain combines them and perceives depth  Trick human visual system  Stereo 3D content production: technical, psychological, and creative skills

3D Video - Issues 10  Not as straightforward as 2D production  Several considerations for a good 3D experience  Balance between 3D effect and overall experience  Minimize viewing discomfort Stereoscopic comfort zone Scene depth adaptation Control of global and local disparity Video composition

Stereoscopic Comfort Zone 11  Comfort zone  Stereographer  “bring the whole real world inside this virtual space”

Control of Absolute Disparity 12  Convergence is controlled by shifting the views

Scene Depth Adaptation 13  Keep in mind disparity range (screen size and resolution)  Carefully plan a scene’s 3D effect  Consider transitions and provide resting periods  Post-production depth adaptation Manual changes per display

3D Display Adaptation 14  From cinema to TV  Depth composition has to modified (stereographer)  Depth information allows for virtual view interpolation

Local Disparity Adaptation 15  Objects should be within stereoscopic window  Intentional depth changes Physical / multiple camera rigs Synthetic / depth-editing  Objects at the border Can cause retinal rivalry Should be cropped by virtually shifting screen plane closer However, not applicable to live broadcast

Live auto-correction 16  Automatic correction & manipulation of stereo live broadcast  Live sport events (big player)  Close objects could abruptly appear  Open research problem  Some kind of novel view synthesis

Video composition 17  Mixing and Composition of 3D Material, Real and Animated Content  Graphics overlay / subtitles  Cannot be simply pasted over other footage  Leverage knowledge about depth range of footage

3D Video - Issues 18  Summary  3D Production has an “art” component  Different medium requires different parameters  Content makers / directors need to think about 3D issues  Stereographer try to balance 3D effect with overall viewing experience

3D video 19  Fundamentals  3D display technologies  Basic processing  View synthesis  2D to 3D conversion  3D media for mobile devices  Outlook

3D Display Technologies 20  “Offer immersive experience”  3D Glasses (cinema + TV)  Head-mounted displays  Volumetric and holographic displays  Autostereoscopic displays

Autostereoscopic displays 21  Best choice for mobile devices  Backward compatible & closer to viewer expectation  Most common Parallax barrier Lenticular sheet

Autostereoscopic displays 22  Crosstalk (ghosting) is most important parameter  “information meant for one eye intrudes into the other eye’s view”

Autostereoscopic displays 23  Issues  Generally less available depth range  More ghosting artifacts  Also, depth information is essential Synthesize different views

3D video 24  Fundamentals  3D display technologies  Basic processing  View synthesis  2D to 3D conversion  3D media for mobile devices  Outlook

Basic processing 25  Signal processing to avoid visual artifacts  Any small visual discrepancy could cause discomfort  Three main categories 1. Correction of Geometrical Distortions 2. Color Matching 3. Adjustment of Stereo Geometry

Correction of Geometrical Distortions 26  Camera rigs might not be perfectly aligned  Real lenses impose radial distortions by nature  Other lens parameters might not sync  E.g. geometrical lens distortions or chromatic aberration

Color Matching 27  Color discrepancy can lead to eyestrain and visual fatigue  Manual calibration need to be done on cameras  Automatic methods exist (histogram filtering)

Color Matching 28  Modern professional postproduction tools incorporate stereo color matching and grading

Adjustment of Stereo Geometry 29  Convergence need to be selected and balanced carefully to achieve good stereo content  Usually by shifting images horizontally in contrary directions, however, cropping & scaling is needed “shift-crop-scale” Demo

Adjustment of Stereo Geometry 30  Stereo baseline is fixed during shooting  Several hardware solutions help camera team analyze the disparity range  Also help visualize result of shift-crop-scale

3D video 31  Fundamentals  3D display technologies  Basic processing  View synthesis  2D to 3D conversion  3D media for mobile devices  Outlook

3D Depth information 32  Depth information is needed for  novel output images in post-production  adjusting the view parallax (different screens)  many different uses in postproduction

Depth info 33  Extracting depth information  (time of flight camera / SfS)  Structure-from Stereo (SfS)  Advanced computer vision problem  Stereo matching Local – block matching, optical flow est. Global – graph cuts, simulated annealing

Structure-from Stereo 34  Global methods are more accurate Slow + don’t work well on video / motion  Local methods are more widely used Window-based methods Some system are in real-time Blocky output

Example of Depth-based method 35  Apply hybrid recursive matching (HRM)  Follow by cross-trilateral median filtering (ACTMF)  Semi-automatic

Post processing depth 36  Align depth discontinues to object borders  Remove noise and mismatches  Fill occlusions  Approaches  Use image segmentation  Neighborhood filtering

View synthesis 37  Synthesize new virtual stereo views by image- based rendering Input – depth + color Output – image with new view  Two types depth-based + warping-based

Depth-based 38  Computer vision techniques Image-based rendering (IBR) Depth-based rendering (DIBR) Layered-depth images (LDI) Intermediate view reconstruction (IVR)

Depth-Image-Based Rendering 39  Need pixel-by-pixel depth maps  Recent focus  Handle depth discontinues  Better depth boundaries

Warping-based 40  Methods that deform the image content directly  Compress or stretch by nonlinear warp function  Do not need camera calibration, segmentation, fill holes  Worst case, visible wobbling artifact

View synthesis 41  Summary Depth maps are computed using computer vision techniques (still active) Generate new views by image-based rendering or warping Warping methods can potentially have less visual problems DEMO

3D video 42  Fundamentals  3D display technologies  Basic processing  View synthesis  2D to 3D conversion  3D media for mobile devices  Outlook

2D to 3D 43  3D to 2D is trivial  Hot topic for 3DTV and 3D cinema  Methods so far are Manual (computer assisted) Automatic

Depth cues 44  Human visual system Monocular cues Binocular cues

Depth cues 45  Monocular depth cues - things can be seen by one eye (2D Camera) Focus / defocus, perspective, relative size Light and shading and texture Motion parallax  Binocular – 2 eyes or 3D camera rigs accommodation, convergence, and binocular discrepancy

Manual 2D to 3D 46  Applicable to prerecorded video  Utilize depth cues to generate a stereoscopic view for each frame  Time consuming and costly  Cost vs. quality

Manual 2D to 3D 47  Three major steps Rotoscoping / segmentation Depth assignment Inpainting  Few companies provide process as a service

Depth Assignment 48  Shifting different parts of scene to simulate 3D  To avoid cardboard effect hire a “3D compositor” Create displacement maps for each pixel Use 3D primitives, spheres or cubes Use DIBR or 3D warping to synthesize view cannot handle transparencies well

Occlusion filling 49  Also known as “in-painting”  One of the most challenging parts in 2D/3D conversion

Automatic conversion 50  Automatic systems Extract depth information + synthesize stereoscopic images

Automatic segmentation 51  Automatic segmentation of background Background subtraction techniques (ML) Optical-flow-based (uses motion as cue)  Objects in scene segmentation Contour-based (color, edges, texture) “SnapCut” state of the art algorithm in After Effects

Automatic depth estimation 52  3D Structure recovery (Shape From X)  Depth from focus/defocus  Depth from geometric cues Relative sizes, gradients  Depth from color and intensity Lights and shadow  Depth from motion Motion parallax

Automatic in-painting 53  Relay on - texture synthesis or border continuation  Some work is done on spatio-temporal inpainting Enforce global temporal consistency for patches  Object motion could reveal background Not viable in practice

Real-time and Offline conversion 54  Real-time Approximate using motion parallax Could use color + intensity information Height-depth cues Hybrid approaches are better  Offline Apply structure from motion Create virtual scene using different views Slow cloudification?

2D to 3D 55  Summary  Hire specialist to rebuild depth and frame objects  Relay on computer vision methods to build depth info  Fill background using in-painting  Still an active research topic segmentation + depth estimation + synthesis

3D video 56  Fundamentals  3D display technologies  Basic processing  View synthesis  2D to 3D conversion  3D media for mobile devices  Outlook

3D media for mobile devices 57  Consider size of display  Computing power + Network constraints  Encoding 3D Video coding (V+D) Multiview Video Coding (MVC)

Issues 58  Stereoscopic images have different disparity  Coding artifacts Dealing with different channels  Transmission artifacts Image distortion on loss Temporal mismatch  Display artifacts autostereoscopic display problems (ghosting)

3D media for mobile devices 59  Not only mobile devices, Internet connected devices  Standards for 3D video encoding are still developing Challenges for 2D video on mobile and more (depth)  3D Video conferencing demand real-time processing  Display technologies are evolving Resolution, need an adaptive and future ready format

3D video 60  Fundamentals  3D display technologies  Basic processing  View synthesis  2D to 3D conversion  3D media for mobile devices  Outlook

Outlook 61  Today, 3D video technology are becoming practical  Content creation process have matured still a lot of room for improvement and extension  There is a huge commercial incentive in this field  A lot of areas to work on…

Outlook 62

Commentary 63  3D has been historically about generating revenue using mind tricks  Tricks are getting much better!  Immersive experience ≠ quality

Commentary 64  Future of 3D Video Consumers decide it Good for short media If all devices support it  Cinema products are profitable!

Commentary 65  Similar strategies could apply on different technology  Light-field camera – Lytro first product 2011 (2006)  Nokia / Pelican Imaging 16-lens array camera 2014

Thank you! 66

Video demo 67 