image stitching

63
Image Stitching Shangliang Jiang Kate Harrison

Upload: valin

Post on 06-Jan-2016

41 views

Category:

Documents


0 download

DESCRIPTION

Image Stitching. Shangliang Jiang Kate Harrison. What is image stitching?. What is image stitching?. Introduction. Are you getting the whole picture? Compact Camera FOV = 50 x 35°. Introduction. Are you getting the whole picture? Compact Camera FOV = 50 x 35° - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Image Stitching

Image Stitching

Shangliang JiangKate Harrison

Page 2: Image Stitching

What is image stitching?

Page 3: Image Stitching

What is image stitching?

Page 4: Image Stitching

Introduction

• Are you getting the whole picture?– Compact Camera FOV = 50 x 35°

Page 5: Image Stitching

Introduction

• Are you getting the whole picture?– Compact Camera FOV = 50 x 35°– Human FOV = 200 x 135°

Page 6: Image Stitching

Introduction

• Are you getting the whole picture?– Compact Camera FOV = 50 x 35°– Human FOV = 200 x 135°– Panoramic Mosaic = 360 x 180°

Page 7: Image Stitching

Recognizing Panoramas

• 1D Rotations ()– Ordering matching images

Page 8: Image Stitching

Recognizing Panoramas

• 1D Rotations ()– Ordering matching images

Page 9: Image Stitching

Recognizing Panoramas

• 1D Rotations ()– Ordering matching images

Page 10: Image Stitching

Recognizing Panoramas

• 2D Rotations (, )– Ordering matching images

• 1D Rotations ()– Ordering matching images

Page 11: Image Stitching

Recognizing Panoramas

• 1D Rotations ()– Ordering matching images

• 2D Rotations (, )– Ordering matching images

Page 12: Image Stitching

Recognizing Panoramas

• 1D Rotations ()– Ordering matching images

• 2D Rotations (, )– Ordering matching images

Page 13: Image Stitching

Recognizing Panoramas

Page 14: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 15: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 16: Image Stitching

Overview

• Feature Matching– SIFT Features– Nearest Neighbor Matching

• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 17: Image Stitching

SIFT Features

• SIFT features are…– Geometrically invariant to similarity transforms,

• some robustness to affine change

– Photometrically invariant to affine changes in intensity

Page 18: Image Stitching

Overview

• Feature Matching– SIFT Features– Nearest Neighbor Matching

• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 19: Image Stitching

Nearest Neighbor Matching

• Find k nearest neighbors for each feature– k number of overlapping images (we use k = 4)

• Use k-d tree– k-d tree recursively bi-partitions data at mean in the

dimension of maximum variance– Approximate nearest neighbors found in O(nlogn)

Page 20: Image Stitching

Overview

• Feature Matching– SIFT Features– Nearest Neighbor Matching

• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 21: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 22: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 23: Image Stitching

Overview

• Feature Matching• Image Matching

– Random Sample Consensus (RANSAC) for Homography

– Probabilistic model for verification

• Bundle Adjustment• Image Compositing• Conclusions

Page 24: Image Stitching

Overview

• Feature Matching• Image Matching

– Random Sample Consensus (RANSAC) for Homography

– Probabilistic model for verification

• Bundle Adjustment• Image Compositing• Conclusions

Page 25: Image Stitching

RANSAC for Homography

Page 26: Image Stitching

RANSAC for Homography

Page 27: Image Stitching

RANSAC for Homography

Page 28: Image Stitching

RANSAC for Homography

Page 29: Image Stitching

RANSAC for Homography

Page 30: Image Stitching

Overview

• Feature Matching• Image Matching

– Random Sample Consensus (RANSAC) for Homography

– Probabilistic model for verification

• Bundle Adjustment• Image Compositing• Conclusions

Page 31: Image Stitching

Probabilistic model for verification

Page 32: Image Stitching

Finding the panoramas

Page 33: Image Stitching

Finding the panoramas

Page 34: Image Stitching

Finding the panoramas

Page 35: Image Stitching

Overview

• Feature Matching• Image Matching

– RANSAC for Homography– Probabilistic model for verification

• Bundle Adjustment• Image Compositing• Conclusions

Page 36: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 37: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 38: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment

– Error function

• Image Compositing• Conclusions

Page 39: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment

– Error function

• Image Compositing• Conclusions

Page 40: Image Stitching

Bundle Adjustment

• New images initialised with rotation, focal length of best matching image

Page 41: Image Stitching

Bundle Adjustment

• New images initialised with rotation, focal length of best matching image

Page 42: Image Stitching

Error function

• Sum of squared projection errors

– n = #images– I(i) = set of image matches to image i– F(i, j) = set of feature matches between images i,j

– rijk = residual of kth feature match between images i,j

• Robust error function

Page 43: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment

– Error function

• Image Compositing• Conclusions

Page 44: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 45: Image Stitching
Page 46: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 47: Image Stitching

Blending

Page 48: Image Stitching

Gain compensation

Page 49: Image Stitching

How do we blend?

Linear blending Multi-band blending

Page 50: Image Stitching

Multi-band Blending

• Burt & Adelson 1983– Blend frequency bands over range

Page 51: Image Stitching

Low frequency ( > 2 pixels)

High frequency ( < 2 pixels)

2-band Blending

Page 52: Image Stitching

3-band blendingBand 1: high frequencies

Page 53: Image Stitching

3-band blendingBand 2: mid-range frequencies

Page 54: Image Stitching

3-band blendingBand 3: low frequencies

Page 55: Image Stitching

Panorama straightening

Heuristic: people tend to shoot pictures in a certain way

Page 56: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 57: Image Stitching

Overview

• Feature Matching• Image Matching• Bundle Adjustment• Image Compositing• Conclusions

Page 58: Image Stitching

Conclusion

Page 59: Image Stitching

Algorithm

Page 60: Image Stitching

AutoStitch program

AutoStitch.net

Page 61: Image Stitching

Open questions

• Advanced camera modeling– radial distortion, camera motion, scene motion,

vignetting, exposure, high dynamic range, flash …

• Full 3D case – recognizing 3D objects/scenes in unordered datasets

Page 62: Image Stitching

Credits

• Automatic Panoramic Image Stitching Using Invariant Features, 2007– Matthew Brown and David G. Lowe (Uni. of British

Columbia)

• Recognising Panoramas, 2003– Matthew Brown and David G. Lowe (Uni. of British

Columbia)– 2003– Thanks for the slides!

• Image Alignment and Stitching: A Tutorial, 2006– Richard Szeliski (Microsoft)

Page 63: Image Stitching

Questions?