I wonder whether image sharpening is actually as effective as it’s portrayed in media like forensics shows. For example, could you sharpen a blurry picture of someone’s face and get a recognizable result? The output on the slide does look sharper, but it also seems a lot noisier than the original. I feel like small details could still be difficult to make out, even in the sharpened image.
charshou
Could sharpening the image too much possibly lead to aliasing since you are adding higher frequencies back into the image?
theflyingpie
From what I have seen, sharpening images can sometimes lead to some aliasing or unwanted artifacts -- for instance in images that my smartphone automatically sharpens, sometimes I will see excessively bright or excessively dark pixelated rims around the edges of objects.
brianqch
@charshou, I think you're right about that! Looking into it a bit further, I think that when images are sharpened, we are enhancing edges and details which means we are enhancing these higher frequency components. Because of this, it seems like excessive sharpening would result in more noise and artifacts.
KevinXu02
Sharpen operation actually makes edges more obvious, but it will also affect those area should be smooth and introduce artifacts like more noice and aliasing.
Mehvix
@charshou Sharpening corresponds to high-pass filtering the image -- this isn't 'adding' high frequencies to the image; rather, it takes away the low frequencies.
Mehvix
@brianqch is correct that high-pass filtering may enhance noise/artifacts (as these tend to be small, pixel-size imperfections i.e. shot noise) -- importantly, this is not caused by aliasing (there is no sampling)
marilynjoyce
Is it enhancing higher frequencies or ridding of lower frequencies? I think ridding lower frequencies, right? Seems like there’s more edge detection.
I wonder whether image sharpening is actually as effective as it’s portrayed in media like forensics shows. For example, could you sharpen a blurry picture of someone’s face and get a recognizable result? The output on the slide does look sharper, but it also seems a lot noisier than the original. I feel like small details could still be difficult to make out, even in the sharpened image.
Could sharpening the image too much possibly lead to aliasing since you are adding higher frequencies back into the image?
From what I have seen, sharpening images can sometimes lead to some aliasing or unwanted artifacts -- for instance in images that my smartphone automatically sharpens, sometimes I will see excessively bright or excessively dark pixelated rims around the edges of objects.
@charshou, I think you're right about that! Looking into it a bit further, I think that when images are sharpened, we are enhancing edges and details which means we are enhancing these higher frequency components. Because of this, it seems like excessive sharpening would result in more noise and artifacts.
Sharpen operation actually makes edges more obvious, but it will also affect those area should be smooth and introduce artifacts like more noice and aliasing.
@charshou Sharpening corresponds to high-pass filtering the image -- this isn't 'adding' high frequencies to the image; rather, it takes away the low frequencies.
@brianqch is correct that high-pass filtering may enhance noise/artifacts (as these tend to be small, pixel-size imperfections i.e. shot noise) -- importantly, this is not caused by aliasing (there is no sampling)
Is it enhancing higher frequencies or ridding of lower frequencies? I think ridding lower frequencies, right? Seems like there’s more edge detection.