This part of the project was particularly interesting to me because there isn't anything particularly technically difficult or anything different than the previous projects we've done in this class, but I learned the importance of having enough data; for days I struggled with the code assuming that there was an issue with my homographies and in the end learned that I simply did not have enough data for the images to turn out well. I also had some trouble with blending working properly but I ended with a decent enough solution that had some artifacting but overall it turned out pretty decently, and I was pretty satisfied. I think the issues I had with the first input/output set are likely because of subjects being both in the foreground and background of the picture; in order to account for both of these, I would have had to input a significantly higher number of correspondence points, which I tried but didn't succeed as well as I would have liked with.
Input 1 | Input 2 | Output image |
---|---|---|