For this part of the project, I took the rectified images of these lego figurines. At first, I grabbed an average of all of the images in the dataset and it turned out to look like this:
After doing so, I essentially used the given X and Y coordinates from each photo and shifted each photo to that (X, Y) coordinate. I also would center the shift around (8, 8), which I had chose because the photoset had 17 x 17 images. I ended up choosing a weight range of [1, 4] to scale my images by on the X, Y shift. I tried a wide range of scalings such as [-2, 2], [-1, 3], and etc. but it seemed like [1, 4] produced the best images. You can tell that the front of the photo starts off more clear and the back walls of the legos become more clear after. Here is my gif and a couple of pictures of my legos:
For this part of the project I essentially started with a small sample over a perpendicular grid and iterated until I ended up at a large sample over a perpendicular grid. This mimics both a small aperture and a large aperture in a camera. I chose my center again to be about (8, 8) and widened the length of the grid. I then shifted the images and averaged them to produce this particular aperture change over my legos:
As someone without any photography background or knowledge, this project really introduced me to the different portions of a camera and more in-depth knowledge about cameras like refocusing and apertures.