2.6 Software
2.6.2 Image Processing
In Chapter 5 we outline how we analysed very large sequences of images in order to produce an accurate particle count throughout the sequence. In order to isolate the particles reasonably the images were recorded using dark field microscopy. This meant that all particles were lighter than the background, making background sub- traction of the image simpler. Image J was used to do the image analysis, using a Macro with a for loop to run through the entire sequence.
Initially the image was converted to an 8-bit image, as it makes image processing more straightforward than having a RGB image, and the image does not have a colour component. In the sequences analysed, any colour in the image is unimportant for the particle analysis. Then the background was subtracted from the image where a dark background is assumed, and a rolling ball radius of 15 pixels was used (the scale of the image is 3.313 pixels/ m, as measured from a sample of gold bands with known dimensions, to calibrate), this defines the smallest particle size which is picked out from the background. If the rolling ball radius is set too low then extraneous noise is counted, and if it is much larger than the particle size, the background
0 255 0 255
(a) i)
(a) ii) (b) i)
(b) i)
Figure 2.9: Grey-scale images (i) and histograms (ii) for images with (b) and without (a) contrast enhancement.
subtraction looks patchy, and large areas of the background are seen as particles. These extremes of having a rolling ball radius which is too low and too high are illustrated in Figure 2.8 a and c respectively. It was sometimes necessary to alter this value throughout the sequence if a particularly low or high supersaturation was used.
Next, the contrast was enhanced, with histogram equalisation. This meant that the histogram of the image took full advantage of the whole grey scale spectrum, and had the maximum possible contrast. This is illustrated in Figure 2.9. This ensured that images throughout the sequence were made more consistent, so that further image processing could be applied to each image, and the particle count produced
(a)
(b)
Figure 2.10: Binary images showing (a) an image of surface nucleation of barite which has been converted to binary and (b) the same image which has had a particle analysis performed which excludes noise and includes holes.
was una↵ected by the original brightness of the image (which often fluctuated due to alterations made to the microscope mid-experiment, and by nucleation material in bulk solution aggregating and obscuring light as it flowed through the cell).
The image was converted to a binary image with a given lower threshold, which was dependent on the size of particles in the image (upper threshold was always set to a maximum of 255). In order to set an appropriate lower threshold throughout the sequence, a cross section of images in the sequence were tested for a suitable lower threshold, and a rough relationship between image number and lower threshold was calculated and applied to the for loop of the Macro.
When the binary image was produced, the particle analysis was performed, this involved removing very small particles (area40 square pixels) and filling particles which contained holes. This is illustrated in Figure 2.10. The parameters that can be calculated as a result of this are inexhaustible, but typically, the area coverage, number of particles and average particle size was printed into a table for further
PLANE-RESOLVED KINETICS OF SALICYLIC ACID
CRYSTALS
3.1
Abstract
The growth and dissolution kinetics of salicylic acid crystals are investigated in
situ by focusing on individual micro-scale crystals. From a combination of optical
microscopy and FEM modelling, it was possible to obtain a detailed quantitative picture of dissolution and growth dynamics for individual crystal faces. The ap- proach uses real-time in situ growth and dissolution data (crystal size and shape
as a function of time) to parameterise a FEM model incorporating surface kinetics and bulk to surface di↵usion, from which concentration distributions and fluxes are obtained directly. It was found that the (001) face showed strong mass transport
(di↵usion) controlled behaviour with an average surface concentration close to the solubility value during growth and dissolution over a wide range of bulk saturation levels. The (¯110) and (110) faces exhibited mixed mass transport/surface controlled behaviour, but with a strong di↵usive component. As crystals became relatively large, they tended to exhibit peculiar hollow structures in the end (001) face, ob- served by interferometry and optical microscopy. Such features have been reported in a number of crystals but there has not been a satisfactory explanation for their origin. The mass transport simulations indicate that there is a large di↵erence in flux across the crystal surface, with high values at the edge of the (001) face com- pared to the centre, and this flux has to be redistributed across the (001) surface. As the crystal grows, the redistribution process evidently can not be maintained so that the edges grow at the expense of the centre, ultimately creating high index internal structures. At later times, we postulate that these high energy faces - starved of material from solution - dissolve and the extra flux of salicylic acid causes the voids to close.