Study of Image Quality of Superimposed Projection Using Multiple Projectors

1. Study of Image Quality of Superimposed ProjectionUsing Multiple Projectors Takayuki Okatani, Member, IEEE, Mikio Wada, and Koichiro Deguchi, Member, IEEE 指導教授 張元翔 報告人員 陳昱辰

2. INTRODUCTION ‧a high-resolution image can be generated by superimposing multiple lower-resolution images projected by different projectors. • analyze the quality of the images realized by this method and show its several fundamental properties.

3. INTRODUCTION ‧A key problem in superimposed projection is the computation of the images fed to the component projectors in order to realize a desired image. ‧we mainly consider the box PSF having the same size as a pixel, which is considered to be a good approximation in the case of a DLP projector.

4. MULTIPROJECTOR SUPER-RESOLUTION • Model of the Image Formation

5. MULTIPROJECTOR SUPER-RESOLUTION • Computing the Projector Images ‧This is subject to the lower and upper bounds of the projector pixel brightness

6. DEPENDENCY OF REALIZATION ACCURACY ON IMAGE STRUCTURES • Analyzing Toy Problems ‧Let us consider the case of 1-D images. We assume the prefilter p and the reconstruction filter r to be box functions whose widths are equivalent to the pixel size of the target image and the projector image,respectively.

7. Fig.1 A single dark pixel

8. Fig.2 A single bright pixel

9. Fig3.4.5

10. DEPENDENCY OF REALIZATION ACCURACY ON IMAGE STRUCTURES ‧Experimental Results ‧The accuracy of image realization by superimposed projection is affected by the target image itself. A typical example is that black-on-white text images tend to be realized more accurately than white-on-black text images.

11. TRADEOFF BETWEEN REALIZATION ACCURACY ANDIMAGE BRIGHTNESS ‧Degrees of Freedom in the Projector Image Optimization Maximum brightness is considered to be an important factor in projection-based-display. the target pixel is less bright, none of these projector pixels may reach the upper bound of the pixel brightness.

12. Fig7.8 • a = 1 a = 5

13. Fig6.9

14. TRADEOFF BETWEEN REALIZATION ACCURACY ANDIMAGE BRIGHTNESS • Experimental Results There is a general tendency that the realization accuracy monotonically deteriorates as the brightness of the target image increases.

15. DEMONSTRATION BY A REAL SYSTEM • Experimental System

16. DEMONSTRATION BY A REAL SYSTEM ‧Experimental methods For the geometric calibration, we used the phase-shift method,which is widely used in range finders, to establish the pixel-to-pixel correspondences between the images of the projector and the calibration camera. Using a sinusoidal brightness pattern for projection, we estimates the phase of the pattern from the brightness changes at each camera pixel when shifting the phase of the sinusoidal pattern.

17. Fig11.12

18. DEMONSTRATION BY A REAL SYSTEM ‧Experimental Results These results verify our theoretical results with respect to the image quality of superimposed projection; they hold in the case of real systems and images.

19. SUMMARY AND DISCUSSIONS • the realization accuracy depends on the structure of the target image to be realized • The method of (multi) projector superimposed projection is itself a novel imaging technology and could have many potential applications not limited to ordinary image projection.