CS194-26 Project 2: Fun with Filters and Frequencies!

Part 1

The first part of the project was about calculating image gradients using various methods. To find the gradient with respect to x and y, we convolved the image with the D_x and D_y arrays shown in the project webpage to get the gradient image along that axis. We then combined the two images using the equation sqrt(df/dx ** 2 + df/dy ** 2). To get the edge image for 1.1, I set all pixels under a certain threshold to zero. In part 1.2, we first blurred the image by convolving a gaussian kernel with the image, then applying the aforementioned process to get results. To speed up this process, we combined the gaussian convolution and gradient convolution into one convolution then found the edge image.

Between the edge images generated with and without blurring, the image generated without blurring had much sharper, disconnected edges. The image generated with blurring was connected throughout, but was much less sharp. The blurring also removed small edges in the grass behind the cameraman which the non blurred edge image couldn't filter out with any threshold.