I built a basic image edge detector. Each image above represents a different step in the pipeline. First the image was smoothed using cross-correlation with a 5x5 Gaussian kernel, effectively smoothing the image. Edges were then enhanced by computing the gradient magnitudes (root sum of squares of X and Y gradients), applying normalization, and converted to black-and-white with a threshold of 120 intensity.
The application was built in Python, and the original image was taken at the Millenium Park in Chicago.