Corner rounding is an inevitable consequence of the current production techniques for photomasks used in optical lithography. The shape of the resulting rounding is dependent on two parameters, beam width and the sensitivity of the photoresist used in production. To simplify detection of faults it would be beneficial to model this rounding with simpler shapes. This project explores two different models for corner rounding, a square with perfectly rounded corners, and a Fernandez-Gausti squircle, two shapes which can be smoothly transitioned between a square and a circle using a roundness parameter. The measure of deviation used for determining the optimal match between the true shape of the corner and the model was the difference between their respective diffraction patterns. Using Fourier optics to analyze the diffraction pattern in the far-field yields the intensity patterns as a scaled Fourier transform of the aperture evaluated at certain frequencies. The diffraction patterns of the apertures were examined both analytically and numerically. While some analytical results were derived, their weak performance relative to the numerical solutions led to the use of exclusively numerical methods for pattern matching. Using the SciPy package for FFT in Python the patterns from the different apertures were calculated and using the least squares method with a logarithmic loss function the optimal roundness parameter that best matched the diffraction pattern from the true shape for a given beam width and sensitivity value could be found. This process was repeated for several combinations of beam width and photoresist sensitivity yield. The beam width yielded the greatest impact on the choice of roundness parameter and shape that produced the best results. A wider beam favored a rounder shape, and specifically the squircle, and vice versa. Some but not a definite effect from the sensitivity of the photoresist could be proven. The main result is that for a given combination of beam width and photoresist sensitivity, the best shape and the optimal roundness parameter for a chosen shape can be provided, alongside the deviation from the true pattern.