MEP export mpeg4 filesize not related to image size?

starosta wrote on 11/23/2012, 9:05 AM

Hi,

I'm exporting a side-by-side stereoscopic movie to MP4 format, using MEP17+HD (v.10.0.12.2(UC1)). It's a bit longer than my usual movie, so I decided to render at less than full HD. I selected 720p size, and the file still came out at more than 1 gigabyte. So then I rendered it at 480x360, and to my surprise, the file came out even LARGER! What's going on?

I checked the video files by playing them. The 480x360 file did in fact look like a lot less resolution, as it should. Why is it so large?

(I did notice that in the export movie dialog box, the predicted size of the file did not go up or down with my selections of more or less pixels/lines in the export video. This seemed odd, but then MEP is "odd" so often...)

I consulted the manual and could find no guidance there.

Has anyone had a similar problem? Does anyone have any ideas?

thanks

Boris

Comments

gandjcarr wrote on 11/23/2012, 12:34 PM

Hi,

Video resolution is by no means the dominant factor in determining the file size of a video especially if you are trying to change the resolution to something lower than it was originally.

The resolution of the original image will play a role in file size but from there on, compression, video bit rate and audio bit rate take over. If you have a high resolution image and you decide to save it as a lower resolution, you need to increase compression and or decrease bit rate. This is a bit of over simplification, but I am not surprised that trying to reduce the video resolution of a high resolution master increased the file size.

Video resolution really just changes how many pixels are displayed when you view it, and that is a very small part of the file sized. The file size is going to be determined by the resolution of the original image plus whatever compression has been added plus any changes to the original audio and video bit rates.

To put it into a little more perspective, if you have a photograph that was taken at a resolution of 1024×768 pixels but decide to print it at 720x480 pixels, your printer will use less ink and the photograph will not look nearly as good but the printer will still have processed a 1024x768 image it just would have printed fewer dots per inch to get that image on paper. The image would not have been processed any faster, in fact it would have been processed slower while the printer calculated how to allocate the ink.

As I said I greatly over simplified how this actually works because it would have taken pages to explain but just know that simply reducing video resolution will not necessarily result in smaller files.

Good Luck,

George

johnebaker wrote on 11/23/2012, 4:23 PM

Prompted by George - and without trying to get too technical this is why a lower resolution video is not significantly smaller in file size than a higher resolution one - in essence a video is a series of still images shown rapidly one after the other just like the old film movies - in reality it is more comlcated than that.

The basics:

Colours in images and video for showing, for example on a TV or computer monitor, are made by combining Red, Green and Blue light in various combinations.

All images/video have what is called Bit Depth - 24 bit colour is fairly standard for most colour images such as jpeg and for most video formats. This means you get a maximum of 16,777,216 possible colours ranging from Black to White with all other colours in between.

There are 8 bits for each of the three component colours (Red, Green and Blue - RGB) (24 / 3 = 8) which, in turn, means each of these three colours has a maximum of 256 steps in shade from no colour - black, to full colour eg Red. (Look up binary counting if you wish to see why you get 256 steps)

In simplistic terms, when compression occurs, the compression algorithm looks for blocks of adjacent pixels and tries to determine if there is a block which is one of the allowed colour combinations.

In a high resolution image it may find a block, for example, of 8 px of the same colour eg blue - instead of adding all eight pixels into the file as (and here I a using words as an analogy) -

'Pixel 1 is blue, pixel 2 is blue, pixel 3 is blue, ' ...... and so on upto pixel 8 is blue

- the algorithm uses a shorthand which says

'the next 8 pixels are blue'

in other words (pun intended ) why use 32 words when I can compress the information into 6 words.

If the same picture is of a lower resolution, the algorithm will find that the blocks of 'same colour' are much less frequent or non existent, because it has fewer pixels to work with, so it has to write more information into the file.

If you get the picture you will now see why a smaller video resolution is not necessarily significantly smaller than a larer resolution video, shorthand cannot be used so effectively in the lower resolution video so more 'words' have to be used to describe the picture

HTH

John

starosta wrote on 11/27/2012, 9:52 AM

I do have a lot of experience working with JPEGs in stills, so I understand the explanation - thanks for reminding me! Still, I am incredulous that a video file could increase in size, when the output pixel count has been reduced by a factor of FIVE! I have a lot to learn about "codecs" and video settings.

So you've motivated me to further explore the advanced video settings, of which there are several screens worth.

I've started another render with the following settings:

I can't access the MEP manual right now because I have the render running, and my searches for an MEP manual online has not succeeded. Thus I'm a bit in the dark about what settings are best (guess I should have consulted manual first, huh?).

Online, I did find this tidbit about setting rates for streaming media on a page at apple.com (in online manual for their Compressor software): Bit Rate slider and field: (Available only with the “Constant at” setting.) Move the slider to set the overall video bit rate for your output media file between the range of 32 kbps and 2048 kbps, or enter a specific number in the accompanying field.

What's interesting is that they recommend a range of bit rate that is far, far lower than the default value in my MEP dialog (namely 10,000kbps). My current render is running at 3000kbps, to create a 400MB file, and I'm curious to see what the quality will be. What kind of bit rate do you select for say HD or 720p to be published on YouTube?

Online, I tried to find information about the other fields in the dialog, but with no success. Maybe the MEP manual will tell me more...

What is HSS rate?

WHAT is CPB size?

What are Frame QPs?

What is HRD?

and look at all these "advanced tab" settings in the "MPEG-4 settings" dialog:

complete gibberish! Do any of you make adjustments to these? Where can I find some information on optimal settings for various publishing channels? (YouTube, BluRay, DVD, etc.)

thanks,

Boris

johnebaker wrote on 11/27/2012, 12:26 PM

For DVD I set the bitrate at maximum and the quality setting to maximum and use the auto adjust if the movie will not fit onto a DVD disk. I do not touch any of the advanced settings.

What kind of bit rate do you select for say HD or 720p to be published on YouTube?

Youtube have recommended settings here - I find it is best to follow these.

HSS = Hypothetical Stream Scheduler

CPB = Coded Picture Buffer

QP = Quantization Parameter

HRD = Hypothetical Reference Decoder

John