Seam Carving
We want to decrease the size of an image, but keep the "important" parts of the image. The general outline is to:
- 1) Define an energy function across the image for each pixel using dynamic programming
- 2) Find a row/column of pixels that have the lowest "energy"
- 3) Delete these pixels
- 4) Go back to #1
Carved Image
Original Image
Failures
All pictures except for the winter house picture were taken by me.
Tilt Shift and Making Miniatures!
We blur parts of the image to make the image look as if it's a miniature scene. The steps to doing so:
- 1) Define a row/column that will remain clear. This is done using ginput to define a start and end to the row/column
- 2) Apply a Gaussian filter to parts of the image.
- 3) Apply more Gaussian filters to parts of the image further away from the clear part of the image.
- 4) Convert the image to an HSV image, and increase saturation by 50%.
The picture at the football game, the picture at Laussane, Switzerland, and the one taken in the streets of Paris were all taken by me. The rest were sourced from the Internet.