Camera Selection:

Images should be obtained with a high quality CCD. The camera I used in my studies was an SBIG-8300M as shown below:



This is a camera typically used for astronomical purposes but also suited my research pretty well. Some of the features it has are: cooling below -10C, a built in desiccant chamber to prevent frosting, a large 3326 x 2504 pixel full frame CCD sensor, and 16-bit capabilities.

However, if high frame rates are needed during image acquisition (anything below 0.1 second exposures), then a camera with an interline CCD or CMOS sensor will be needed instead. 


Light Setup:

For DIC, the ideal light source is one that provides diffuse axial light. Diffuse light will help mitigate glare/reflections and provide a more evenly lit surface. There are devices called "diffuse axial illuminators" that use a beam splitter to provide axial light, but in my experience, these devices seem to cause aberrations when acquiring images and have a difficult size constraint to deal with. Instead, what worked best for me was a simple semi rigid light guide (available from Edmund Optics).



The optics used for DIC is also a highly important consideration. For my experiments, I used a Questar QM100 long distance microscope as shown below:



This setup was very robust because it provided high magnification at extended distances. This allowed other components (such as a coiled induction heater) to be used in conjunction with a compression/tension machine without getting in the way of the optics. The QM100 was also mounted to a tripod which allowed it to be repositioned and used with other compression/tension machines.


Camera Settings and Aquisition:

The only major camera setting that was adjustable on the SBIG was exposure length. In my experience, the overall exposure length does not seem to have much of an effect on results, as long as it is set correctly and the exposure is not so long as to cause large aberrations due to dark current.

For other cameras, sometimes the gain and gamma are adjustable. Both of these should be set to 1. Increasing the gain will also increase image noise. Instead, the light source should be increased; only after the light source has been saturated should the gain be increased. Lastly, the gamma parameter causes the grayscale values to vary in a nonlinear way. This caused the dark frame calibration (discussed in the next section) to work poorly, since it's based upon the idea of superposition.

Lastly, when acquiring images, make sure that the entire spectrum of grayscale values is within the grayscale bounds as much as possible. These limits are 0 to 255 for 8-bit cameras and 0-65535 for 16-bit cameras. This ensures nothing is saturated and all the data is being collected correctly. A typical (16-bit) histogram for a patterned sample is shown below:



These images should also be saved in an uncompressed format, such as a 16-bit tif file. The .jpg format can also be used, but be aware that compression techniques might be utilized with .jpg which will reduce image quality. 


Image Post Processing:

Once images have been obtained, there are several techniques that can be used to increase the signal to noise ratio. These techniques are generally called image calibration techniques and are commonly used in astronomy.

Various techniques with varying levels of sophistication for image calibration exist, but the technique that seemed to work the best, from my experience, was to use a simple dark frame calibration. A dark frame is an image taken at the same exposure length as a regular image, but without any incoming light (i.e. taken with the lens cap on). This image is then subtracted from a regular exposure to help reduce the effects of dark current, which is dependent on exposure length (i.e. larger exposures result in larger aberrations due to dark current). Regular exposures can also be averaged together to help mitigate noise. The process is outlined below:

One problem with this method is that images obtained in the current configuration can sometimes move as the sample undergoes deformation. This means averaging "current" images may not be feasible. Furthermore, the dark frame subtraction is only necessary for longer exposures. In many of my experiments, short exposures (~0.1 sec) were used; dark frame subtraction can typically be omitted for this case. However, image averaging did seem to have an effect for all exposure settings, and should be used whenever possible.