Michael Wu (myw9)

CS 5670: Computer Vision

November 19, 2012

Project 4: Single View Modeling

VRML Animation of �Book Mountains� Painting by Patrick Hughes

I. Overview

The purpose of this project was to create an application to generate 3D texture-mapped models from a single image using single view modeling. The project involved calculating vanishing points and homographic transformations using user-specified reference points and lines. In addition to creating texture maps, the application also generates VRML (Virtual Reality Modeling Language) code which is used to create an interactive 3D model of the image scene.

II.����� Painting: �Book Mountains� by Patrick Hughes

The VRML animation shown above the overview was generated from the �Book Mountains� painting created by artist Patrick Hughes. If you enjoy the painting, please visit and support the artist�s webpage at http://www.patrickhughes.co.uk/paintings.htm.

(1) Source Image (Flickr): http://www.flickr.com/photos/ereshkigal/98423709/

(2) SVM Annotated Image:

SVM Statistics: 28 Points, 6 Lines, and 10 Polygons

Quadrilateral polygons were created for each face of the bookshelf. In addition, multi-sided polygons were used to model the floor and the ceiling. The width and length of each bookshelf section was assumed to be the same (i.e. a square bookshelf). The height of the bookshelf was assumed to be three times its width/length.

(3) VRML Still Image:

(4) Texture Maps:

����� �

Mountain��������������������������������������������������� ������������Bookshelf

(5) VRML Model: Click HERE to view an interactive 3D model of the painting. *Note: you must install a VRML viewer such as the Cortona browser to be able to see the 3D model.

III.��� Photograph: �Tunnel Vision� by Jim Seft

The 3D single view model shown below was generated from a photograph called �Tunnel Vision� by Jim Seft. If you enjoy the photo, please support the photographer at http://www.flickr.com/photos/jimsef/.

(1) Source Image (Flickr): http://www.flickr.com/photos/jimsef/4395107475/

(2) SVM Annotated Image:

SVM Statistics: 35 Points, 14 Lines, and 13 Polygons

Because of the rounded edges on the ceiling of the tunnel, several quadrilateral polygons were used to cover the entire ceiling. The tunnel walls and floor were also primarily modeled using quadrilaterals, except for the left and right corners where triangles were used to approximate the corresponding textures. The end of the tunnel was modeled using a single multi-sided polygon that fits with the shape of the ceiling.

(3) VRML Still Image:

(4) Texture Maps:

�����������������������

End of Tunnel��������������������������������������������������� Floor Tiles

(5) VRML Model: Click HERE to view an interactive 3D model of the tunnel photograph. *Note: you must install a VRML viewer such as the Cortona browser to be able to see the 3D model.