Approach:
For tilt shift, the idea was that we assess the image in rows. There are a few arguments: the center of the focused area, the width of the focued area, the width of the gradient, and the strength of the defocus area. Using this, we start at the center of the focused area, then we assess each row. For the gradient area, we use a gaussian filter to blur the outside. To do this, I created a gaussian stack of the images. I then use samples from this stack in order to generate a smoother filter (rather than just incrementing the sigma value each time)
Source
Source
Approach:
For this shot, what I did was start kind of far away from the object. Then, you physically move closer by a bit, then zoom out a bit. This creates the vertigo shot effect. I found that using backgrounds with a hallway or somethign in the backgrond makes the effect much nicer.
