iago
20th August 2002, 19:35
@vlad59
First of all, thanks for the reply. And I must say that I actually don't see the speed issue so much matter of concern and absolutely prefer to focus on quality, because imho the major purpose of all this respectable effort going on here is what you've already pointed out: quality (keeping enough details as well as removing the noise).
And another point regarding the speed of convolution3d: with matrix1 (1, ...) and applying it after simple resize [just like in the case of Cnr2().temporalSoften(1,5,0) chain], the difference in performance is not really "that" big (sth like -25%) compared to the previous filter chain. Actually, ATM I'm quite pleased with the current encoding speed.
And yes, the best thing to say for now is: "Let's wait for the result." :)
best regards,
iago
First of all, thanks for the reply. And I must say that I actually don't see the speed issue so much matter of concern and absolutely prefer to focus on quality, because imho the major purpose of all this respectable effort going on here is what you've already pointed out: quality (keeping enough details as well as removing the noise).
And another point regarding the speed of convolution3d: with matrix1 (1, ...) and applying it after simple resize [just like in the case of Cnr2().temporalSoften(1,5,0) chain], the difference in performance is not really "that" big (sth like -25%) compared to the previous filter chain. Actually, ATM I'm quite pleased with the current encoding speed.
And yes, the best thing to say for now is: "Let's wait for the result." :)
best regards,
iago