Here we sampled the image at twice the horizontal and vertical resolution. Then we blur that 2x-resolution grid of samples to get rid of the jaggies. Is this different than blurring the jaggies as shown on slide 22? Why or why not?
DannyTran123
I think that blurring the 2x-resolution grid of samples to get rid of the jaggies is different than blurring the jaggies, because when you blur the jaggies the shape of the jaggies is still present. Since the jaggies are still present, I believe that blurring the jaggies will just create blurred jaggies. I was however, a little confused when professor Ng said that a lot of engineering problems would be simplified if we were able to blur the jaggies and then do aliasing because at the end of the day don't we have to do the same operations just in a different order.
Sicheng-Pan
From my perspective blurring the jaggies is actually sampling at 1x resolution and filter to get another 1x resolution signal, which is different from what is being done here.
joeyzhao123
I think this is really to cool to see the end result here and connect it back to video games we've played. I used to sort of thing that my anti-aliasing just made edges a little bit blurry but this adds more clarity to how it works and what the different options mean.
zhaominl
Thanks to Joey's comments. It reminds me of adjusting antialiasing level on Nvidia control panel before (I was not sure about its meaning at that time). I browse this topic and found Nvidia did some interesting research on anti-aliasing. Here is one which developed an algorithm for real-time adaptive supersampling in game link.
rsha256
New question:
What do the 25%, 50%, 75%, 100% mean? Is that how many of the 4 subpixels are within the box and if so then how do we represent that wrt the color we end up drawing -- do we change the alpha channel (transparency) or do we re-weight the overall color (how does that even work with RGB??)?
(Seed for discussion)
Here we sampled the image at twice the horizontal and vertical resolution. Then we blur that 2x-resolution grid of samples to get rid of the jaggies. Is this different than blurring the jaggies as shown on slide 22? Why or why not?
I think that blurring the 2x-resolution grid of samples to get rid of the jaggies is different than blurring the jaggies, because when you blur the jaggies the shape of the jaggies is still present. Since the jaggies are still present, I believe that blurring the jaggies will just create blurred jaggies. I was however, a little confused when professor Ng said that a lot of engineering problems would be simplified if we were able to blur the jaggies and then do aliasing because at the end of the day don't we have to do the same operations just in a different order.
From my perspective blurring the jaggies is actually sampling at 1x resolution and filter to get another 1x resolution signal, which is different from what is being done here.
I think this is really to cool to see the end result here and connect it back to video games we've played. I used to sort of thing that my anti-aliasing just made edges a little bit blurry but this adds more clarity to how it works and what the different options mean.
Thanks to Joey's comments. It reminds me of adjusting antialiasing level on Nvidia control panel before (I was not sure about its meaning at that time). I browse this topic and found Nvidia did some interesting research on anti-aliasing. Here is one which developed an algorithm for real-time adaptive supersampling in game link.
New question:
What do the 25%, 50%, 75%, 100% mean? Is that how many of the 4 subpixels are within the box and if so then how do we represent that wrt the color we end up drawing -- do we change the alpha channel (transparency) or do we re-weight the overall color (how does that even work with RGB??)?