A macro lens can deliver incredibly shallow depth of focus — an effect often seen when photographing very small objects. We can simulate this effect by Gaussian blurring parts of an image, leaving one area unblurred.
In this project, I created a program that allowed me to define an upper
and lower bound for a depth of field (a horizontal stripe of the image).
Then, I created a Gaussian pyramid of l = 5, sigma = 2, as in
Project 3.
I iterated through all the pixels, checking its y-coordinate, and using a hashing function to decide which layer of the Gaussian stack I would sample from to fill in that pixel. The further from the DOF, the blurrier (higher up the pyramid). To get rid of harsh edges, I used a mask to blend with the original image. Since the DOF area should be not blurry, the alpha mask has a low value that linearly increases from that DOF.
Gaussian stack levels
Mask for alpha blending
Original
Miniature
Gaussian stack levels
Mask for alpha blending
Original
Miniature
Gaussian stack levels
Mask for alpha blending
Original
Miniature
Gaussian stack levels
Mask for alpha blending
Original
Miniature
The dolly effect can be recreated by taking photos of a subject at different distances with different focal lengths on a zoom lens. The key is to keep the subject in a constant position and size within the frame. I did this by moving backwards while zooming in on the lens.
An extra step I took was to align all the frames in Photoshop to make the subject as "still" as possible. This helped make the GIF animation much more realistic and appealing.
