Vertigo Shot and Fake Miniatures

CS 194-26 Final Project  ·  Arnav Vaid

Part 1: Vertigo Shot

Theory

The dolly zoom is a cinematography technique used to create a dizzying effect by keeping an object in focus while moving away from it and zooming to keep the proportion the object is in the image. To simulate this effect, I took several pictures of objects while backing away and zooming appropriately. 

Results

beer chair

An issue I encountered was keeping the angle of view of the object constant over images. You can see that the last frame of the beer gif has a slightly different angle than the rest. 

Part 2: Fake Miniatures

Theory

We can make landscape images look like miniature sets by simulating the effect of a limited depth of field. To achieve this, first we let the user specify a line of focus and whether they want a vertical or horizontal effect. Next, I generate a gaussian stack where the sigma value changes by roughly 0.75 every time. From there, I generate a set of masks that mask out slits of the image that are of increasing distance from the selected area of focus. I applied a gaussian filter to these masks to get blurred masks. Finally, I apply the blurred masks to their corresponding layer in the gaussian stack. This makes parts of the image that are further from the focused portion more blurrier. Finally I stitch these masked gaussian stacks together to get the final image. I also applied 50% saturation increase to give a more "fake" effect.

Results

Results were varied depending on the images - I found more complicated images gave worse results since it made the seams of the image more obvious. This is most evident in the picture of Toronto. The first 3 images I found off the internet, while the rest I took during a trip to London.

Toronto

Before After

Village

Before After

Suburb

Before After

London 1

Before After

London 2

Before After
 

London (Cathedral)

Before After

Conclusion

I found the Fake Miniatures project most fun to do. Overall, I'm glad I took this course to understand the basics of image processing and common practices (image pyramids, HSV, Gradients, etc.). Thanks for a great semester!