You are viewing the course site for a past offering of this course. The current offering may be found here.
Lecture 22: Image Processing (31)
nebster100

I don't really see a difference between this and the standard box blur. Can someone help me see it?

EmmmaTao

I think the difference is the weights assigned to the "box" - for Gaussian blur, the weight are not uniform but weighted by the neighboring pixel's relative distance to the pixel under consideration. For box blur, the weights are uniform, which just means taking the average of the neighboring 7*7 pixel values. Therefore, visually the gaussian blur preserves the basic structure of the bricks better.

julialuo

@nebster100 if you mean the visual difference between the two, notice that with the Gaussian blur you can still trace the black outlines of the bricks while with the box blur if you just saw the zoomed in version you wouldn't really even recognize the bricks. This is because the Gaussian blur puts a higher weight on the actual color of the center pixel, which Emma discusses above. Also, notice that the box blur has strange aliasing effects (random high frequency vertical lines appearing throughout). I believe this is because the box blur in the frequency domain actually appears to selectively allow some high frequencies through, while the Gaussian blur is still a Gaussian (no bias) in the frequency domain. You can see these translations to the frequency domain back in lecture 3 (this is the one for box blur: https://cs184.eecs.berkeley.edu/sp19/lecture/3-95/antialiasing).

chenwnicole

That's a good point on how the blurs differ visually! Tying it back to lecture 3 is pretty cool, thanks for the reference. I definitely noticed how the black outlines seemed different in the two blurs, but couldn't pinpoint how they differed.

You must be enrolled in the course to comment