Image-based Lighting Design

# Image-based Lighting Design

## Image-based Lighting Design

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
##### Presentation Transcript

1. Image-based Lighting Design 8 Sept. 2004 Frederik Anrys Philip Dutré Computer Graphics Group

2. Light intensities are determined by using the photograph. Overview Paint a design on a photograph of a real life object.

3. Previous Work • Lighting Design • Sketching highlights and shadows to position lights (Poulin ‘97) • Painting with Light (Schoeneman ‘93) • Image-based Relighting • A lighting reproduction approach to live-action composition (Debevec ‘02)

4. Overview

5. Overview

6. Overview

7. Overview

8. Light Stage (Debevec) • 40 IColor MR RGB Led Lights • Simulate the visible spectrum once calibrated Acquisition: Lights

9. Acquisition: Camera • Canon EOS D30 digital Camera • High Dynamic Range (HDR) Images • Luminance values are represented in full floating point values instead of 8-bit RGB values used for display. • No compression using a response curve. • Usually constructed by taking a series of photographs

10. Overview

11. Basis images • For each light source, construct a HDR image. • Linearize each HDR image into a vector . • Store each vector into matrix A.

12. Design Specification • Start painting desired lighting design using Adobe Photoshop. • Adjust color, contrast, add shadows, … • Linearize it into vector Y.

13. Overview

14. Optimization Find x such that y A is minimal

15. Optimization • Ax resides in luminance space • y resides in display space • F converts Ax to display space • Why? The L2 norm is well defined and fast method for comparing images in display space. • F is standard Gamma correction function

16. Optimization • Objective function is non-linear. • Constrained in • added for favoring regions/pixels

17. Optimization • The problem is solved by Sequential Quadratic Programming (SQP). • Gradient is computed analytically. • Optimization takes 15 seconds to complete: • 40 lightsources (3 channels). • P3-1.2 Ghz computer.

18. Target Result Target Result Target Result Results

19. Results

20. Result Result Target Target Results

21. Future Work • Determining light positions/directions. • Multiple camera viewpoints. • Give designer interactive feedback.

22. Thank you all. Questions? Acknowledgements • Graphics group from K.U.Leuven. • Artists for lending out their work.

23. Virtual Applying configuration to Light Stage: Reality check Target Real