Augmented and Mixed Reality

Similar documents
Multi-View Stereo for Static and Dynamic Scenes

Multiple View Geometry

Stereo. Outline. Multiple views 3/29/2017. Thurs Mar 30 Kristen Grauman UT Austin. Multi-view geometry, matching, invariant features, stereo vision

Multiview Reconstruction

Dense 3D Reconstruction. Christiano Gava

CS4495/6495 Introduction to Computer Vision. 3B-L3 Stereo correspondence

Dense 3D Reconstruction. Christiano Gava

But, vision technology falls short. and so does graphics. Image Based Rendering. Ray. Constant radiance. time is fixed. 3D position 2D direction

Fundamentals of Stereo Vision Michael Bleyer LVA Stereo Vision

CS 4495 Computer Vision A. Bobick. Motion and Optic Flow. Stereo Matching

EE795: Computer Vision and Intelligent Systems

A Low Power, High Throughput, Fully Event-Based Stereo System: Supplementary Documentation

Overview. Augmented reality and applications Marker-based augmented reality. Camera model. Binary markers Textured planar markers

Stereo Vision. MAN-522 Computer Vision

Image Based Rendering. D.A. Forsyth, with slides from John Hart

Three-Dimensional Sensors Lecture 2: Projected-Light Depth Cameras

MERGING POINT CLOUDS FROM MULTIPLE KINECTS. Nishant Rai 13th July, 2016 CARIS Lab University of British Columbia

Multiple View Geometry

Static Scene Reconstruction

Outline. ETN-FPI Training School on Plenoptic Sensing

EE795: Computer Vision and Intelligent Systems

Occlusion Detection of Real Objects using Contour Based Stereo Matching

Final project bits and pieces

CS 4495 Computer Vision A. Bobick. Motion and Optic Flow. Stereo Matching

Image-Based Modeling and Rendering

Stereo vision. Many slides adapted from Steve Seitz

Hybrid Rendering for Collaborative, Immersive Virtual Environments

Volumetric Scene Reconstruction from Multiple Views

Multi-view stereo. Many slides adapted from S. Seitz

Computer Vision Lecture 17

Step-by-Step Model Buidling

Computer Vision Lecture 17

Modeling, Combining, and Rendering Dynamic Real-World Events From Image Sequences

Chaplin, Modern Times, 1936

Video Mosaics for Virtual Environments, R. Szeliski. Review by: Christopher Rasmussen

CSCI 5980: Assignment #3 Homography

Ping Tan. Simon Fraser University

Image-Based Rendering

Depth from two cameras: stereopsis

CS5670: Computer Vision

Vision-Based Registration for Augmented Reality with Integration of Arbitrary Multiple Planes

The Video Z-buffer: A Concept for Facilitating Monoscopic Image Compression by exploiting the 3-D Stereoscopic Depth map

Light source estimation using feature points from specular highlights and cast shadows

Topics to be Covered in the Rest of the Semester. CSci 4968 and 6270 Computational Vision Lecture 15 Overview of Remainder of the Semester

Stereo Vision A simple system. Dr. Gerhard Roth Winter 2012

Range Imaging Through Triangulation. Range Imaging Through Triangulation. Range Imaging Through Triangulation. Range Imaging Through Triangulation

AIT Inline Computational Imaging: Geometric calibration and image rectification

Stereo. 11/02/2012 CS129, Brown James Hays. Slides by Kristen Grauman

Augmented Reality, Advanced SLAM, Applications

Visualization 2D-to-3D Photo Rendering for 3D Displays

Approach to Minimize Errors in Synthesized. Abstract. A new paradigm, the minimization of errors in synthesized images, is

Recent Trend for Visual Media Synthesis and Analysis

Fast Natural Feature Tracking for Mobile Augmented Reality Applications

COMP 102: Computers and Computing

Machine vision. Summary # 11: Stereo vision and epipolar geometry. u l = λx. v l = λy

Recap: Features and filters. Recap: Grouping & fitting. Now: Multiple views 10/29/2008. Epipolar geometry & stereo vision. Why multiple views?

EECS 442 Computer vision. Stereo systems. Stereo vision Rectification Correspondence problem Active stereo vision systems

Direct Plane Tracking in Stereo Images for Mobile Navigation

Shadows. COMP 575/770 Spring 2013

Segmentation and Tracking of Partial Planar Templates

Stereo: Disparity and Matching

Stereo and structured light

Epipolar Geometry Prof. D. Stricker. With slides from A. Zisserman, S. Lazebnik, Seitz

3D Visualization through Planar Pattern Based Augmented Reality

Subpixel accurate refinement of disparity maps using stereo correspondences

Challenges and solutions for real-time immersive video communication

REFINEMENT OF COLORED MOBILE MAPPING DATA USING INTENSITY IMAGES

Outdoor Scene Reconstruction from Multiple Image Sequences Captured by a Hand-held Video Camera

The Light Field and Image-Based Rendering

Outline. Introduction System Overview Camera Calibration Marker Tracking Pose Estimation of Markers Conclusion. Media IC & System Lab Po-Chen Wu 2

Adaptive Multi-Stage 2D Image Motion Field Estimation

PANORAMIC IMAGE MATCHING BY COMBINING HARRIS WITH SIFT DESCRIPTOR

Colorado School of Mines. Computer Vision. Professor William Hoff Dept of Electrical Engineering &Computer Science.

Lecture 15: Image-Based Rendering and the Light Field. Kayvon Fatahalian CMU : Graphics and Imaging Architectures (Fall 2011)

A Comparison between Active and Passive 3D Vision Sensors: BumblebeeXB3 and Microsoft Kinect

Textureless Layers CMU-RI-TR Qifa Ke, Simon Baker, and Takeo Kanade

LS-ACTS 1.0 USER MANUAL

Lecture 6 Stereo Systems Multi-view geometry

There are many cues in monocular vision which suggests that vision in stereo starts very early from two similar 2D images. Lets see a few...

Real-Time Video-Based Rendering from Multiple Cameras

CHAPTER 3 DISPARITY AND DEPTH MAP COMPUTATION

3D Editing System for Captured Real Scenes

Rendering and Modeling of Transparent Objects. Minglun Gong Dept. of CS, Memorial Univ.

Motion Tracking and Event Understanding in Video Sequences

Dynamic Spatial Partitioning for Real-Time Visibility Determination. Joshua Shagam Computer Science

Acquisition and Visualization of Colored 3D Objects

A virtual tour of free viewpoint rendering

Using temporal seeding to constrain the disparity search range in stereo matching

Tecnologie per la ricostruzione di modelli 3D da immagini. Marco Callieri ISTI-CNR, Pisa, Italy

But First: Multi-View Projective Geometry

Stereo and Epipolar geometry

Invariance of l and the Conic Dual to Circular Points C

On-line and Off-line 3D Reconstruction for Crisis Management Applications

Neue Verfahren der Bildverarbeitung auch zur Erfassung von Schäden in Abwasserkanälen?

3D from Images - Assisted Modeling, Photogrammetry. Marco Callieri ISTI-CNR, Pisa, Italy

Omni-directional Multi-baseline Stereo without Similarity Measures

Colorado School of Mines. Computer Vision. Professor William Hoff Dept of Electrical Engineering &Computer Science.

12/3/2009. What is Computer Vision? Applications. Application: Assisted driving Pedestrian and car detection. Application: Improving online search

WATERMARKING FOR LIGHT FIELD RENDERING 1

Lecture 9 & 10: Stereo Vision

Transcription:

Augmented and Mixed Reality Uma Mudenagudi Dept. of Computer Science and Engineering, Indian Institute of Technology Delhi

Outline Introduction to Augmented Reality(AR) and Mixed Reality(MR) A Typical AR System Issues in AR Different methods of Augmenting object Algorithms in AR Abstract Model of AR/MR Initial Results Obtained Summary

Introduction to AR and MR AR combines real and virtual objects in real environment Virtual objects(3d model/images/video) are merged with real environment Milgram et al(1994) described the relation between AR, MR and VR REAL ENV Mixed Reality VR AR AV Figure 1: Continuum of real and virtual environment MR spectrum lies between the extremes of real life and Virtual reality(vr). Views of the real world are combined in some proportion with views of a virtual environment

Typical AR system AR SYS = Computer vision + Computer graphics + User interfaces SCENE CO ORD CAMERA POSITION REAL SCENE PZT REAL IMAGE CO ORD WORLD CO ORD VIR OBJ CO ORD VIR OBJ ALIGN GRAPHICS CAMERA TO REAL GRAPHICS RENDERING GRAPHICS CO ORD VIDEO IMAGE GRAPHICS IMAGE GRAPHICS IMAGE CO ORD Figure 2: Typical AR system

Issues in AR:Registration Process of estimating an optimal transformation between two images(also known as spatial Normalization) To align the virtual object to real objects in 3D

Issues in AR:Registration contd.. Most critical requirement of AR system :Since human visual system is very good at detecting even small mis-registration Methods: No generalized method of registration for all the type of augmentations Static errors:optical distortion, mechanical misalignment and incorrect viewing parameters Dynamic errors:system delays

Issues in AR: Tracking Tracking : View point tracking as the view point moves Tracked viewing pose defines the AR alignment and registration Issues: Foreshortening, Scaling, Occlusions

Issues in AR: Modeling Modeling : Modeling of the destination with Texture and extraction of the 3D model from the source environment Single view : Single view reconstruction Multiple view reconstruction: two view

Possible ways of augmenting Source 3D Model: Model is given Single Image: Extract 3D/2D Model of the object and Texture Multiple Images: Extract 3D Model of the object and Texture Video : Extract 3D Model of the object and Texture Destination 3D Model:Model of destination Single Image: Single view Reconstruction with Texture Multiple Images: Multiple view reconstruction with Texture Video: Tracking and Registration of destination

Possible ways of augmenting Source 3D Model: Model is given Single Image: Extract 3D/2D Model of the object and Texture Multiple Images: Extract 3D Model of the object and Texture Video : Extract 3D Model of the object and Texture Destination 3D Model:Model of destination Single Image: Single view Reconstruction with Texture Multiple Images: Multiple view reconstruction with Texture Video: Tracking and Registration of destination

Possible ways of augmenting Source 3D Model: Model is given Single Image: Extract 3D/2D Model of the object and Texture Multiple Images: Extract 3D Model of the object and Texture Video : Extract 3D Model of the object and Texture Destination 3D Model:Model of destination Single Image: Single view Reconstruction with Texture Multiple Images: Multiple view reconstruction with Texture Video: Tracking and Registration of destination

Source:3D Model, Destination:Single Image Destination Image Augmented image

Possible ways of augmenting Source 3D Model: Model is given Single Image: Extract 3D/2D Model of the object and Texture Multiple Images: Extract 3D Model of the object and Texture Video : Extract 3D Model of the object and Texture Destination 3D Model:Model of destination Single Image: Single view Reconstruction with Texture Multiple Images: Multiple view reconstruction with Texture Video: Tracking and Registration of destination

Source:Multiple Images, Destination:Single Image Source Images(2/7) Augmented image

Possible ways of augmenting Source 3D Model: Model is given Single Image: Extract 3D/2D Model of the object and Texture Multiple Images: Extract 3D Model of the object and Texture Video : Extract 3D Model of the object and Texture Destination 3D Model:Model of destination Single Image: Single view Reconstruction with Texture Multiple Images: Multiple view reconstruction with Texture Video: Tracking and Registration of destination

Source:Multiple Images, Destination:Multiple Images Source Images(2/20) Augmented image

Possible ways of augmenting Source 3D Model: Model is given Single Image: Extract 3D/2D Model of the object and Texture Multiple Images: Extract 3D Model of the object and Texture Video : Extract 3D Model of the object and Texture Destination 3D Model:Model of destination Single Image: Single view Reconstruction with Texture Multiple Images: Multiple view reconstruction with Texture Video: Tracking and Registration of destination

Source:3D model, Destination:Video Source1, Augmented1 Source2, Augmented2, Augmented2, Augmented2 and Augmented2 Augmented3

Possible ways of augmenting Source 3D Model: Model is given Single Image: Extract 3D/2D Model of the object and Texture Multiple Images: Extract 3D Model of the object and Texture Video : Extract 3D Model of the object and Texture Destination 3D Model:Model of destination Single Image: Single view Reconstruction with Texture Multiple Images: Multiple view reconstruction with Texture Video: Tracking and Registration of destination

Mixed Reality Mixed Reality Example Microsoft research lab: SIGGRAPH 2004 Video view interpolation using layered approach Color Segmentation based stereo algorithm Mattes near discontinuities Two layer compressed representation to handle matting Major disadvantages:synchronizing many cameras, and acquiring and storing of images

Move matching method Source:3D Model, Destination:Video Given by Zisserman et.al Initialization: Manually indicate the planar region in image-0. (corner detection and matching are restricted to this region) Detect interest points in image 0 Initialize camera calibration K Steady state: computing H from frame i to i 1 Detect interest points in two images say X 1 k N 1 k 1 Match interest points which maximizes the cross correlation in 7x7 mask: x j x 1 k X j N j 1 and

Move matching method contd.. Randomly sample subset of four matched pairs and compute homography. Each candidate H is tested against all the correspondence by computing distance between x 1 and Hx. Choose H for which most pairs are within the threshold Compute pose from H i Result set 1: Source, Tracking and Augmented Result set 2: Source Tracking and Augmented 1 i

Method-2,G. Simon et.al Results Augmented Initialization stage Camera parameters 3D/2D corr of 4 pts. computation of initial pose + set of visible model features in the first image + extraction of key points in the first frame Model of the obj to be added in the scene. Trajectory of the object in the scene image k >k+1 Tracking of set of visible model features in the current image Updating the set of tracked pts STEP 1 Key points are extracted in frame k+1. They are matched with the points extracted in k STEP 2 The view point is computed using mixing method STEP 3 The computer generated object is added in scene STEP 4

Augmented views:z-keying method z-keying method: Given by T. Kanade et.al Uses dense depth map as a switch For each pixel, the z-key switch compares the pixel depth values of two images, and routes the color value of the foreground image that is nearer to the camera for the merged output image Real and virtual objects will occlude correctly Uses real time stereo-machine:specifications are Number of cameras: 2 to 6 Frame rate: max 30frames/sec Depth image size: up to 256 240 Disparity search range:60 pixels Results: Augmented

Abstract Model of AR/MR Static Part 3D model of space Dynamic Parts Temporal Info (real+virtual) Query F(V,t) 4D data: Space and Time Render Image

Problem of 4D Fly-through Static 3D Model: Reconstructed from set of images Temporal Information: Extract Temporal information for each view 4D data:3d space and time Query:F V t? Generate view from the static and temporal information and answer the query

Results Source:3D model and Destination: Still image Method: Single view reconstruction method Destination Image 3D model augmented into still image Augmented movie-1, Augmented movie-2 and Augmented movie-3

Results contd.. Source:Object extracted from video and Destination:Still image Method: Source: Object is extracted as 2D plane Destination: Sparse 3D modeling of possible occluders Registration: Object plane to ground(virtual) plane Homography Rendering: Modified ray tracing algorithm source movie Augmented movie

summary Typical AR system and Issues in AR Possible augmentation methods and review results Abstract model of AR and MR 4D fly through problem Preliminary results to wards the 4D fly through problem

Thank you THANK YOU