A Virtual Telescope on Mobile DevicesQiyuan Tian, Qingyi Meng

Department of Electrical Engineering, Stanford University

Mobile Image Matching Application Feature-based Matching (SIFT/ SURF)

Speeded Up Robust Features (SURF)

[Bay et al., ECCV 2006]

Scale-invariant feature transform (SIFT)

[Lowe, ICCV 1999]

Mobile Virtual Telescope System

Query

Information

Wireless

Network

Reference�D.G. Lowe, "Distinctive image features from scale-invariant

keypoints", Int. J. Comput. Vision, 60 (2) (2004), pp. 91 –110.

�H. Bay, T. Tuytelaars, L.V. Gool, SURF: speeded up robust

features, in: European Conference on Computer Vision, vol. 1,

2006, pp. 404–417.

�J.M. Morel and G.Yu, ASIFT, A new framework for fully affine

invariant image comparison. SIAM Journal on Imaging Sciences,

2(2):438-469 (2009)

Client Motorola MOTA855

Network 802.11gn

Server Lenovo Laptop Y470

2.3 GHz 4 GB RAM + WAMP

Database 8 Stanford Buildings

System Description

Zoom in

Camera

Capture

Image

Matched +

Satellite

Images

Crop

Image

Match

Features

Amplified

Object

Extract

Features

Display

Viewfinder

KDTree RANSAC

xaaa

aaax

=

121110

020100'

(1124, 1072) SIFT features

1124 pre-RANSAC matches

11 post-RANSAC matches

ANN Affine Model

Threshold

Feature-based Matching (Affine-SIFT)

Acknowledgements: Images and figures used in this poster

courtesy of corresponding authors.

Future Work�User Interface

�Low-resolution Image

retrieval (multiple zooming in

and retrieval/ more robust

features for small images)

� Affine invariant Image

matching

�Geographic information to

assist retrieval (Client: GPS

+ inertial sensor; server:

images with location tag)Image Formation Model Affine-SIFT in One FigureMain Decomposition

SIFT:

�Rotation and translation are

normalized.

�Zoom is simulated in scale space

�No latitude and longitude

Affine-SIFT:

�Simulate latitude and longitude,

then apply SIFT

),(*),(isqsq DDmeanThreshDDd <