I fear I'm getting ill. Headaches, exhaustion. All around I'm just not feeling very well. I realize that has little to do with the project, but I wanted to write it somewhere.
At first this last portion seemed very easy to implement. Instead of drawing a red line over the seam, shift pixels over to cover the seam and return the given image -1 width. It worked. But for larger images was taking far, far too long. I looked into seam-carving algorithms and discovered my method of calculating the seam was inefficient. Once updated the issue went away, and things appear to be working fine.
I admit the code could be cleaner. The lack of polish is a sacrifice I took to make sure the project gets done in a reasonable time.
That's all for now. I need some rest. Hopefully I feel better, soon.
Ciao,
-Abigail
This one's easy since we made code for a vertical seam already. Just transpose the image along its diagonal, run the vertical seam, then transpose it again.
This took a bit longer than I wanted to. Events in my personal life postponed progress. Everything was simple enough, though. I haven't had to touch shortest path algorithms in a while, so it was nice to get a refresher.
I accidentally weighted the path on intensity instead of energy, which caused a problem, but otherwise everything turned out alright.
Once again, I did not read the question thoroughly and caused myself unnecessary woe.
In the function BufferedImage.energyOf() I check the passed X,Y values to make sure they are on the border.
I then have adjustX and adjustY for border pixels.
Then, when calculating the bordering pixels, leftPixel, rightPixel, topPixel, bottomPixel I used the adjusted values for Both X and Y
My code looked like this:
val leftPixel = Color(this.getRGB(adjustX - 1, adjustY* ))
val rightPixel = Color(this.getRGB(adjustX + 1, adjustY* ))
val topPixel = Color(this.getRGB(adjustX*, adjustY - 1 ))
val bottomPixel = Color(this.getRGB(adjustX*, adjustY + 1 ))
\\ *this is where the problem is, in case you don't see it.
What that did was unnecessarily shift what values I was evaluating away from the border. Basically, I was ignoring a 1px trim around the entire image.
From the description
Energy for border pixels is calculated with a 1-pixel shift from the border. For example:
E(0,2)=Δ2x(1,2)+Δ2y(0,2)−−−−−−−−−−−−−−−√
E(1,0)=Δ2x(1,0)+Δ2y(1,1)−−−−−−−−−−−−−−−√
E(0,0)=Δ2x(1,0)+Δ2y(0,1)−−−−−−−−−−−−−−−√
But what I was doing was
Energy for border pixels is calculated with a 1-pixel shift from the border. For example:
E(0,2)=Δ2x(1,2)+Δ2y(1,2)−−−−−−−−−−−−−−−√
E(1,0)=Δ2x(1,1)+Δ2y(1,1)−−−−−−−−−−−−−−−√
E(0,0)=Δ2x(1,1)+Δ2y(1,1)−−−−−−−−−−−−−−−√
So. I was getting errors that the Hash didn't match, but when I looked at the output images and compared to the example, everything SEEMED fine. I even opened GIMP to check the difference of the two images. I didn't find a thing, but I"m guessing the difference was just very subtle.
Alas! A very small oversight made a very large nuisance! But I guess that's just how coding goes sometimes. Measure twice cut once, etc. I look forward to learning this lesson again, someday.
Ok. fun. Actually doing stuff with images again. Having to handle parameters. Fun stuff.
I ran into a silly issue with this. IDEA said there was no main found. It's because I had used fun main(args: List<String>) instead of fun main(args: Array<String>)
Oops! Other than that though, Easy enough.
This was the usual Stage 1 simple setup. "Create a program that does very little towards the intended goal of the project."
It was a nice refresher with working with BufferedImage and ImageIO, which I hadn't touched in a few weeks.