the_corona
15th August 2010, 17:28
Given enough time and a defined frame x, or a sequence of frames 1-n, is it possible to calculate the minimum bitrate required to achieve a given quality (whatever measure, doesn't matter) for a given quality.
Or put differently, what is the maximum compressibility of a defined frame (or sequence) for a given codec (spec). Let's keep psych-enhancements out of the picture.
It is my understanding that encoders basically take all sorts of shortcuts to try to approach this optimum. They have to do this in order to keep encoding times at a sane level.
Is this one of those famous NP-complete problems, that is, are there simply too many variations that need to be tried to proof that a given result is the optimum within a given spec? My instinct tells me it is (though it is probably feasible for a single or even a small number of frames).
Assuming my assumption is correct (lol), can we apply a approximation/statistical model to say that we are n% sure a bitstream is within m% of the theoretical optimum?
Are such models used in encoding research?
Or put differently, what is the maximum compressibility of a defined frame (or sequence) for a given codec (spec). Let's keep psych-enhancements out of the picture.
It is my understanding that encoders basically take all sorts of shortcuts to try to approach this optimum. They have to do this in order to keep encoding times at a sane level.
Is this one of those famous NP-complete problems, that is, are there simply too many variations that need to be tried to proof that a given result is the optimum within a given spec? My instinct tells me it is (though it is probably feasible for a single or even a small number of frames).
Assuming my assumption is correct (lol), can we apply a approximation/statistical model to say that we are n% sure a bitstream is within m% of the theoretical optimum?
Are such models used in encoding research?