I noticed that the contrast in the sharpened image increased. This intuitively makes sense because when performing sharpening we are putting more weight on the pixel's own value and reducing the value by the fall-off between the chosen pixel and neighboring pixels, thus enhancing the difference between adjacent pixels with different colors, but is this always the desired effect?
starptr
There's quite a bit of artifacts in the zoomed-in sharpened image. Maybe it's possible to get rid of the "random" pixels by introducing a blur as well?
rsha256
What happens when we repeatedly sharpen an image? Surely there is a limit to how much we can sharpen -- does it have no effect after a certain point or does it distort the image?
If so then how do we tell the optimal amount of sharpening without distorting?
AndyyHuang
I wonder what happens if you apply a blur and then sharpen the image. Would it be possible to get a less aliased final image due to the speckles on each brick being blurred out before sharpened.?
Noppapon
After some thoughts about the question @AndyyHuang posted, I reckon sharpening an image after a blur can never recover back the original image. This is because the inverse of, say, a box blur will never be in the form of a sharpen matrix.
StaffDanCubed
I also wonder if attempting to antialias the sharpened image will just remove the sharpening effect. After all, I think that the more dynamic and clear image we see is the direct result of the pixels looking this way when zoomed in, so maybe if we try to blur-then-sharpen, it won't even look very sharp anymore.
rsha256
I wonder if ML could help learn how to correct images here
I noticed that the contrast in the sharpened image increased. This intuitively makes sense because when performing sharpening we are putting more weight on the pixel's own value and reducing the value by the fall-off between the chosen pixel and neighboring pixels, thus enhancing the difference between adjacent pixels with different colors, but is this always the desired effect?
There's quite a bit of artifacts in the zoomed-in sharpened image. Maybe it's possible to get rid of the "random" pixels by introducing a blur as well?
What happens when we repeatedly sharpen an image? Surely there is a limit to how much we can sharpen -- does it have no effect after a certain point or does it distort the image? If so then how do we tell the optimal amount of sharpening without distorting?
I wonder what happens if you apply a blur and then sharpen the image. Would it be possible to get a less aliased final image due to the speckles on each brick being blurred out before sharpened.?
After some thoughts about the question @AndyyHuang posted, I reckon sharpening an image after a blur can never recover back the original image. This is because the inverse of, say, a box blur will never be in the form of a sharpen matrix.
I also wonder if attempting to antialias the sharpened image will just remove the sharpening effect. After all, I think that the more dynamic and clear image we see is the direct result of the pixels looking this way when zoomed in, so maybe if we try to blur-then-sharpen, it won't even look very sharp anymore.
I wonder if ML could help learn how to correct images here