Why is the distant circle OK in a severly downsampled version, while there is aliasing in full resolution of the image. Does blurring add to the feeling of texture or subtract away (allowing users to nitpick fine details)?
KevinXu02
I believe downsampling filters out high-frequency information from the image, as we discussed in the previous class. For distant objects, our sampling frequency for textures is lower, and downsampling helps avoid signal aliasing, thereby preventing the occurrence of jagged edges. However, for objects closer to us, where the sampling frequency is higher, the downsampling of textures at this point appears to lack detail.
Rohan1215
Different levels of downsampling seem to have a very significant impact on the amount of aliasing in an image. Would there be a way to scan a level 0 image in one pass to calculate of how much detail or high frequency information is present? Could this be used to estimate which level to render an image if the frames are updating quickly?
angelajyzhang
I think the effect of downsampling is so interesting depending on the distance of different objects (i.e. small background objects vs. foreground subjects). That may be why the wall in the OK circle looks fine at this downsampling rate while the ledge at the front is extremely aliased.
Why is the distant circle OK in a severly downsampled version, while there is aliasing in full resolution of the image. Does blurring add to the feeling of texture or subtract away (allowing users to nitpick fine details)?
I believe downsampling filters out high-frequency information from the image, as we discussed in the previous class. For distant objects, our sampling frequency for textures is lower, and downsampling helps avoid signal aliasing, thereby preventing the occurrence of jagged edges. However, for objects closer to us, where the sampling frequency is higher, the downsampling of textures at this point appears to lack detail.
Different levels of downsampling seem to have a very significant impact on the amount of aliasing in an image. Would there be a way to scan a level 0 image in one pass to calculate of how much detail or high frequency information is present? Could this be used to estimate which level to render an image if the frames are updating quickly?
I think the effect of downsampling is so interesting depending on the distance of different objects (i.e. small background objects vs. foreground subjects). That may be why the wall in the OK circle looks fine at this downsampling rate while the ledge at the front is extremely aliased.