Hacker News Reader

Comment by jharger | Hacker News Reader

I remember implementing this for a class years ago, and then the professor suggested doing the inverse to try to expand the image width. The idea was you would duplicate the lowest energy seam... but all that did was create a lot of repeats of the same seam.

I never did finish that weird idea, but I probably needed to try something like increasing the energy of the chosen seam (and its duplicate)... I may try that again, just because I'm curious what would happen.

View on HN

The original seam carving paper discussed expanding too: https://youtu.be/6NcIJXTlugc?t=56

http://www.faculty.idc.ac.il/arik/SCWeb/imret/imret.pdf

> Figure 8: Seam insertion: finding and inserting the optimum seam on an enlarged image will most likely insert the same seam again and again as in (b). Inserting the seams in order of removal (c) achieves the desired 50% enlargement (d). Using two steps of seam insertions of 50% in (f) achieves better results than scaling (e). In (g), a close view of the seams inserted to expand figure 6 is shown.

That's actually done in the paper! You basically choose the lowest energy seam, duplicate it, then blacklist it and duplicate the next lowest energy seam (to prevent repeatedly duplicating the same seam). The results are quite good.

The approach from the original paper is to remove seams like you are decreasing the size, which provides you with a set of exclusive seams that can be added to the original image (with some mapping logic). This produces an output without repeating the same seam repeatedly.

Reply on news.ycombinator.com