redfordxx
22nd March 2005, 20:17
Hi,
I read something about DCT and I hope I got something a little. It's a process to convert the binary information of 8x8 pix block to math expresion ... weighted sum of many cos(..) expressions, correct? The name Fourier comes to me (misty memory from uni)...
Quantization is a process, when the cosines with smallest weight are forgotten, right? (Like mp3?) Correct me if I am wrong. Quantization matrix defines somehow (don't know how), how small the weights (coefficients) have to be, to be allowed forgotten (=0).
And NOW: In 2nd pass there are some other quantizers (1-31). So, does it mean that, that the "cosines serie" produced by DCT is quantized again? Or, what is the relation of quantizers I see in the VDM status window and the quantizers in the quatization matrix?
Side question: Packed bitsrteam --- yes or no?
I read different FAQ and threads with different statements for reasons I cannot remember all now. I don't know, which sould I trust more (I expect to buy standalone, if that's relevant).
I read something about DCT and I hope I got something a little. It's a process to convert the binary information of 8x8 pix block to math expresion ... weighted sum of many cos(..) expressions, correct? The name Fourier comes to me (misty memory from uni)...
Quantization is a process, when the cosines with smallest weight are forgotten, right? (Like mp3?) Correct me if I am wrong. Quantization matrix defines somehow (don't know how), how small the weights (coefficients) have to be, to be allowed forgotten (=0).
And NOW: In 2nd pass there are some other quantizers (1-31). So, does it mean that, that the "cosines serie" produced by DCT is quantized again? Or, what is the relation of quantizers I see in the VDM status window and the quantizers in the quatization matrix?
Side question: Packed bitsrteam --- yes or no?
I read different FAQ and threads with different statements for reasons I cannot remember all now. I don't know, which sould I trust more (I expect to buy standalone, if that's relevant).