An image is just a grid of numbers. Detect an edge with a hand-built filter, count objects with a flood fill, then track motion across frames.
3 units, about 3 hours. Free.
Convolve a hand-built kernel over an image and explain why the response peaks at an edge.
2Segment an image by thresholding and count the separate objects using connected components.
3Detect and track a moving object using frame differencing and centroids.