Getting the Most out of Your Video Gadget Storage - Let's Encode!
(Page 4 of 5 )
As promised earlier, we're going to encode and discuss the results, so that we can figure out a pattern and consequently phrase some conclusions. An article that does not provide any real-world results or practical applications cannot be considered useful enough. We won't make that mistake, so read on!
First of all, our source/original video file is going to be an MPEG2 recorded movie clip accompanied by MPEG2-Audio as its audio codec. The resolution of the source file is 720x576 on 25 fps. Its audio segment comprises 48 kHz on 160 kbps. The file size is approximately 105 MB. The computer used for video transcoding is an Intel Core2 E6300 powered system with 1 GB of memory and runs on Windows 2003 Server.
Now that you know what we're going to work with, let me tell you our objective. Since we are going to do some comprehensive analysis on the potential gains and losses while transcoding, and on which settings are worth changing and playing around with, we should be doing dozens of tests on different settings. Then we can compare!
The output format is going to be 3GPP [.3gp], since that's the standard for GSM-based cell phones. You can opt for 3GPP2 [.3g2] if you own a CDMA2000-based phone. There shouldn't be any difference in the outcomes, since the encoding scheme that we'll use is the MPEG4, regardless of the resulting file format.
We'll convert the source clip into 176x144 (QCIF) and 320x240 (QVGA) resolutions, while playing around with the following transcoding options: FPS (frames-per-second), video bit-rate, audio codec, audio bit-rate, audio frequency. To simplify the comparison, the output filenames are renamed to clearly point out the settings that were used.
Check out the attached table below. The first column contains the filenames, while the second represents the file sizes. Keep in mind that the last row is the original source clip (.mpeg extension, as you will notice). Pay close attention to the table...
Okay - we're back now. What could you deduce from the comparison table above? First of all, the single most critical factor when transcoding multimedia files is the bit-rate. This is what defines how many bits comprise one second (thus, measured on a per-second basis). If you remember, we've discussed audio bit-rates during the first half of this series. The same applies for video media files.
Obviously, the higher the better guideline cannot fail, but we want to maximize our gains, while reducing the losses (without sacrificing quality). Therefore, there is no reason to choose the highest possible setting because the resulting file sizes end up quite large. Basically, chances are that we haven't gained anything from encoding.
The second important factor is the audio. Whether you are transcoding video clips of your favorite artists performing on a scene with lots of effects and fireworks, a short clip where your kid rides the bicycle for the first time, or a trailer of that movie you've been anxiously waiting for, we cannot afford to totally ignore the audio quality of our media files. We'll find a sweet combination that fulfills our needs!
Since the media file I'm converting was a song performed live on a stage by the group t.A.T.u., sacrificing audio quality wasn't an option. I've tried starting out from 12kbps audio bit rates and going up to 128kbps in relatively small steps, while gradually progressing the audio sampling rate from 8,000 Hz up to 48,000 Hz. Keep in mind that 44 kHz is the sampling rate of audio CDs, while 48 kHz is for DVDs. Going higher is pointless in our case because that area should be used only for high definition content.
Let's drift back to the comparison table presented earlier and determine which options do not affect the size of the resulting file that much. First, the frames per second; there's no point saving a few dozen kilobytes when we're losing ten frames each second. Keep your FPS the same as your source clip. In our case that's 25 fps.
Second, we can see that going higher than 64 kbps for the audio bit-rate doesn't affect the quality or the file size that much. The sound quality was, simply put, crap on lower bit-rates than 64 kbps (such as 12-48 kbps). The sampling rate of 24 kHz was also a necessity. Below that, the quality of the song took a huge hit. Thus, we could conclude that the best combination is 64 kbps on 24 kHz.
But there's no better way to figure this out than by doing lots of tests. As you can see, that's what I've done too. You can learn the most by doing it yourself and watching the transcoded clips and noticing the differences. Finally, the most influential factor that defined both the quality of the video frames as well as the filesize was the video bit-rate. I've tried a few variations on 190 kbps and 400 kbps.
As you would expect, increasing this affects the file size almost linearly. We can interpret this as how many bits per second are pumped into the file, so it makes sense. Therefore, 190 kbps was decent, but 400 kbps was twice as good - and most importantly, we could see the difference with our eyes. To exemplify this, I made two screen shots. Check out the attached image and pay attention to the pixelation.
Not everybody will notice the difference, but trust me, there is one. [It's especially noticeable around the eyes. --Ed.] And the easiest way to notice this is by actually transferring both of the files to your portable gadget, cell phone, or media player and then trying to play them back. That's when you are going to notice the difference, when you natively play it back on its display size.
Last but definitely not least, going higher with frame rates per second is pointless. You won't gain anything out of it. As per using a different audio codec from AAC, such as the AMR, the results weren't impressive at all. AMR wasn't designed for music files and these kinds of conversions. Please refer back to the theory presented earlier in this article to understand AMR's scope and definition.
Finally, I've gone with the following settings: 3GP+AAC at 25fps with 400 kbps video bit-rate, 64kbps audio bit-rate, and 24 kHz sampling rate. The resulting file was only about 11.3 MB. Therefore, we have reduced the filesize of the original clip by more than 9 times. That's one hell of a difference. And no, we haven't sacrificed its video or audio quality. We've just tailored the clip to our requirements.
Next: Concluding Thoughts >>
More Hardware Guides Articles
More By Barzan "Tony" Antal
