Overview
When we want to resize an image, we have several options – rescaling, cropping. Rescaling allows us to preserve the aspect ratio as well as relative information about the image (we don't cut anything out), while cropping allows us to reshape the image by cutting out sections we no longer want. What if we want to reshape the image and cut out redundant sections, but we don't want to simply cut away the edges of the image? Seam carving allows us to resize the image in a content-aware way, slicing out pixels that contribute less to the detail or information given by the image.
Method
The first step of any seam carving algorithm is defining an energy function, basically a function that maps pixels
to their level of importance. For our energy function we used a variation of the dual gradient energy function, which basically
takes the difference in pixel values for each channel between the x-neighbor and the y-neighbor and computes the sum of squares.
This is basically edge detection, so pixels that are along an edge will have higher energy and will be preserved compared to pixels
that are not along an edge.
