cbr vs vbr

[tenders]

when trying to encode a movie, which is the best setting,cbr or vbr?

[MITSUI_1]

MPEG tutorial here:

CBR = Constant Bit Rate

This means the encoder is going to spit out an specific, exact (or damn close) rate of bits for every second of video. But video has fluctuating demands. What this means is that during processing, the encoder has to constantly vary the Quantization level to get the data to fit in the fixed bit rate. Quantization is the process whereby the encoder throws away data (details, color, etc) that it cannot "afford" to keep in order to meet the encoding bitrate. The higher the Quantization level, the more stuff is getting thrown away. At an extreme, the encoder throws away ALL of the data for a 16x16 macroblock area eacept for a single color attribute, which means you see a "block" there on the screen instead of the data you want. Macroblocks showing are therefore a symptom of bitrate starvation. Alternatively, there is also a problem with CBR where the set bit rate is actually more than the encoder needs to encode the simple parts of the video stream, even when no information is thrown away by the Quantizer (e.g., the credits). The result is the encoder fills out the extra space with "padding" in order to maintain the constant bitrate. This is obviously a waste of CDR space and is inefficient. Because the Quantization factor is forever fluctuating, the CBR rate must be relatively high or the video will vary from a very good picture in low motion/detail scenes to uttery horrid picture in fast/complex scenes. For VCD at 352x240 and SVCD at 480x480 the "standard" CBR settings of 1150 and 2576 are generally NOT good at getting the best quality. You will usually see macroblocks and compression artifacts during demanding footage.




VBR = variable bit rate

In this approach, the encoder alters the number of bits per second based on the characteristics of the video. It is muchos better. I don't know enough about how multi-pass works to explain it intelligently, but single pass VBR is my favorite and I can go into that. Simply put, a single-pass VBR implementation will usually try to maintain a constant or near constant quantization level. Here it is important to distinguish between "Quantization" and "Quality" because different encoders refer to the different concepts. CCE, for example, uses a "Quanitzation" setting in its one pass VBR. This means that if you want higher quality (i.e., less information thrown away) you set the Quantization level LOWER. TMPGenc, on the other hand, talks in terms of Quality. In its implementation, raising the "Quality" number has the same effect as lowering the Quantization factor in CCE. The one pass implementations will always purposefully throw away just enough information to hover at the quantization level you have chosen. Anyway, the way this works is that when the encoder hits a complex or high motion scene, it will increase the number of bits it throws at the problem in order to keep from throwing away more information than your set Quantization level would allow. The only exception is when you have set a "Max" bitrate. In that case, the encoder will try to increase bitrate to maintain the chosen quantization right up until it hits the max bitrate, at which time the encoder will have no choice but to start increasing its quantization and throw more and more data away. The nice thing about one pass VBR is that as long as you set the max bitrate high enough, the encoder will produce a very consistent file. The downside is that because bitrate is all over the place depending on the needs of the quantizer, the final file size is damn near impossible to predict.

Like I said, I think one pass VBR is the best setting, but its a pain because you will often have to reencode if you get the file size wrong on your first try.

[kinneera]

If you want to fit a certain amount of video in a certain space and maintain the highest quality possible, multiple-pass VBR will always be the superior option, since you can specify an average bitrate to guarantee a final filesize while taking advantage of the more intelligent bit allocation used in VBR. The only exception is if the average bitrate is greater than or equal to the highest allowable bitrate on your playback device (for most standalone DVD players ~2500 Kbps), in which case single pass VBR will be preferable.