Would something like off-grid sampling really be that effective? Because at the end of the day, or at least in this case, if you were to tilt the grid space around or the object we're capturing, wouldn't you get the same results?
Staffyirenng
@DannyTran123 Good point, and you are right that such a periodic sampling as shown here would still be susceptible to aliasing. And you correctly observe that if we rotate the scene to match a rotation in the sampling pattern, we would get exactly the same aliasing patterns as original. (Though, the pattern here is not purely a rotation of the 2x2 supersampling grid. But the general effect you described is quite true.)
All this said, such an off-grid pattern does seem to help in practice, for real-world scenes that we want to render. Perhaps you can theorize why? One hint may be to consider the relative commonality of vertical or horizontal lines in natural imagery.
Would something like off-grid sampling really be that effective? Because at the end of the day, or at least in this case, if you were to tilt the grid space around or the object we're capturing, wouldn't you get the same results?
@DannyTran123 Good point, and you are right that such a periodic sampling as shown here would still be susceptible to aliasing. And you correctly observe that if we rotate the scene to match a rotation in the sampling pattern, we would get exactly the same aliasing patterns as original. (Though, the pattern here is not purely a rotation of the 2x2 supersampling grid. But the general effect you described is quite true.)
All this said, such an off-grid pattern does seem to help in practice, for real-world scenes that we want to render. Perhaps you can theorize why? One hint may be to consider the relative commonality of vertical or horizontal lines in natural imagery.