| The amount of data provided by fluorescing microscopy is often vast, resulting in tedious visual inspection and is frequently subject to human estimation. For the specific case presented here, the study of metastasizing medulloblastoma cancer cells adhering to leptomeningeal cells requires a certain specificity of boundary localization that struggles with human repeatability, noise and changing conditions, large data sets, and manual input. Edge detection is a favorable image processing technique due to its ease of use, computational efficiency and localization.;This thesis presents novel modifications to the Canny edge detection algorithm to better enhance data of fluorescing microscopy in a custom designed imaging system. This algorithm results in enhanced cell localization and boundary specificity as well as alleviating a significant amount of manual input.;The Canny approach is a highly regarded means of edge detection yet it is a global solution that is still susceptible to noise and lacks autonomy. This thesis presents a modification to the Canny method by exchanging the initial smoothing function, providing a novel edge detection operator, and using the popular Otsu threshold method to automatically determine the final threshold values. The work here presents the medical and image processing motivations, an in-depth analysis of the Canny algorithm with proposed modifications, compares results of the traditional Canny method with those of the modified method on the selected fluorescing microscopy, and is followed by our conclusion and desired future work. |
