Module 5: Schlieren and Shadowgraph
  Lecture 29: Review of optical techniques for imaging crystal growth
 

IMAGE PROCESSING

The three optical methods generate images that are path integrals of the refractive index fields and in turn, the concentration (or temperature) distribution in the fluid medium. The integrals can be simplified if the fields being studied are strictly two-dimensional. In general, the local information can be extracted using principles of tomography. In the present work, the images have been interpreted as carrying information of the solutal concentration (or temperature) field that is an average along the direction of propagation of light.

The determination of the concentration/temperature field from interferograms requires several intermediate steps, including (a) noise removal, (b) edge detection, (c) location of intensity minima within fringe bands, and (d) fringe thinning. Step (d) involves fitting a smooth function through points of intensity minima within a single fringe, as obtained in step (c). In addition, assigning temperatures to fringes, followed by transferring the data to a Cartesian grid are important.  The fact that fringe thickness is small in regions of high concentration or temperature gradients has not been used in the present work. The analysis and interpretation of fringe patterns in interferometry has been discussed in detail by various authors. The approach described in has been implemented in the present study.

In contrast to interferometry where information is localized at the fringes, a schlieren image carries information related to the local temperature/concentration in the form of an intensity distribution. The advantage here is that data is available at the pixel-level of the camera. Drawbacks include the errors due to superimposed noise associated with scattering and the possibility of device saturation. In the present work, the first factor is taken to be less significant because the field variable is obtained by integrating the intensity distribution (see Data Reduction), an operation that tends to smooth noisy profiles. The second factor was circumvented by reducing the laser intensity using neutral density filters. The local temperatures (and concentrations) were then determined by numerically integrating the appropriate governing equation in the direction normal to the knife edge.