Does it matter which two points we pick to do the bilinear filtering? Maybe there's an advantage on the closest two points or the furthest?

anthonyhsyu

It seems like we defined the lerp in 1D, so I think we would want the two points to be on the same axis. It should also be the case that if you picked the points that are on the same horizontal axis, we would get the same result.

v-wangg

Based on this, it seems like bilinear filtering applied a type of linear blurring to the pixels similar to the box filter from previous lectures which takes the average of the four nearest neighbours, except with bilinear filtering we add weights s and t to the values of these neighbours according to their respective distances from the sample point. Not sure if this is correct but the intuition seems right?

frgalvan

^jcebreros, I'm pretty late on this, but i think the nearest ones means the nearest reference-able points... In the case of Assignment 1, that would mean the pixel points...? I could be wrong.

Does it matter which two points we pick to do the bilinear filtering? Maybe there's an advantage on the closest two points or the furthest?

It seems like we defined the lerp in 1D, so I think we would want the two points to be on the same axis. It should also be the case that if you picked the points that are on the same horizontal axis, we would get the same result.

Based on this, it seems like bilinear filtering applied a type of linear blurring to the pixels similar to the box filter from previous lectures which takes the average of the four nearest neighbours, except with bilinear filtering we add weights s and t to the values of these neighbours according to their respective distances from the sample point. Not sure if this is correct but the intuition seems right?

^jcebreros, I'm pretty late on this, but i think the nearest ones means the nearest reference-able points... In the case of Assignment 1, that would mean the pixel points...? I could be wrong.