Professional Documents
Culture Documents
Automatic Measurement of Newton Fringe Paper
Automatic Measurement of Newton Fringe Paper
Abstract
Automatic vision-based measurement of Newton fringes on reflective LCOS microdisplays is a challenging problem due to the large variation in the way fringes manifest themselves on a display. The fringes generally appear as low contrast transitions between bright and dark rings or bars. Variations in intensity, spacing, and curvature add to the difficulty of recognizing and characterizing the fringes. In addition, no standard currently exists for counting fringes. Different counting methods are used by various microdisplay companies. This paper attempts to open up a discussion on how to best characterize Newton fringes. An image-processing technique will be presented for processing Newton fringe images. A standard for counting fringes will be proposed. A measure of relative fringe severity will also be proposed.
1. Introduction
Reflective liquid-crystal-on-silicon (LCOS) microdisplays are made by sandwiching liquid crystal material between a silicon backplane and a cover glass. Newton fringes are caused by multiple reflections of light off the glass surfaces and the reflective substrate. These reflections cause constructive and destructive interference patterns that result in alternating bright and dark rings or fringes when viewing the display. Poor flatness in the substrate and thickness variation in the glass can increase the fringe count and intensity of the fringes. The intensity variation of the fringes becomes more acute when the light source is more monochromatic. Because of the poor quality than results when viewing the display, it is important to measure and characterize Newton fringes. A standard way to measure Newton fringes is to simply count the number of fringes on the display. For an automated measurement system, fringes are difficult to measure due to the variations in intensity, spacing, and curvature of the fringes. There is also the question of how to count fringes. No standard currently exists for counting fringes. This paper proposes some techniques for measuring Newton fringes.
Figure 2. Image of Newton fringes on microdisplay #2. the inner white ring. If these rings were complete, the count would be 1.5 at a minimum. If only whole rings were counted, the fringe count would be 0.5. Some manufacturers may say that these rings are nearly complete and thus give the display a count of 1.5. How would one devise an image-processing system to make the same analysis? Figure 3 shows another fringe image. The ring counting method completely breaks down for this case.
Figure 4. Image from Figure 1 processed to segment the image into black and white fringe regions. black and white fringe regions. One could then use standard binary image processing techniques such as connectivity analysis to count the number of black and white regions. However, image-processing techniques based on segmenting fringes are almost impossible to implement due to incomplete fringes and areas of very low contrast. Figure 4 shows one such approach. The Figure 1 image is segmented into black and white regions using an adaptive thresholding process based on gray-level morphology. This process works fine on the outer rings but does not segment out the inner fringes. This leads to a lower fringe count than expected. Another potential approach is to transform the image into the frequency domain using a two-dimensional FFT, for example. However, in most cases, the spectral information for the fringes is so spread out and in the noise that it is impossible to get accurate fringe count.
Figure 3. Image of Newton fringes on Microdisplay #3. One potential image-processing technique for counting the fringes is to segment the image into
CT = Max(T 1, , TN )
CR =
CT 1 . 4
Figure 7. Image of Newton fringes in microdisplay #4. An obvious choice for a severity measurement is a function of the brightness amplitude between the peaks and valleys in the gridline profile. The following measure is proposed:
Figure 6. Profile of upper-left to lowerright gridline in Figure 4. This technique is advantageous for production testing with a vision system since a) it is a good indicator for the cosmetic quality of a microdisplay, b) it is a much less subjective measure of fringe counting, and c) it is fairly straightforward to implement an image processing algorithm for this counting technique. Researchers developing microdisplays may still need to use the ring counting method to get a better understanding of what is causing the fringing to occur. To convert CT to CR, the fringe count using the ring counting method,
% Severity =
i =1
Pi Vi Pi 100 N
where N is the number of the transitions in the gridline with the maximum number of transitions, Pi is the brightness level for the peak in the ith transition in that gridline, and Vi is the brightness level for the valley in the ith transition in that gridline. This formula is a measure of the average peak-to-valley amplitude as a percent of the average brightness of the display. For the image in Figure 1, the severity was measured to be 3.87%. For the image in Figure 7, the severity was measured to be 1.77%. Of course,
one has to be careful when comparing microdisplays using this measure. For this measure to be useful, the spectral characteristics of the light source must be the same for all measurements. In general, a wider spectrum bandwidth will generate smaller severity numbers for a given microdisplay.
5. Conclusion
Two Newton fringe measures have been proposed to characterize the Newton fringes in a microdisplay. A fringe count based on bright-todark transitions and dark-to-bright transitions was proposed as a standard method for counting fringes for production testing of microdisplays. Also a severity measure has been proposed as a complementary measure to fringe count. It is the authors intent that this paper will to a discussion on standard Newton fringe measurements.