0

I am using the following script to convert my MTS video to an AVI video:

for %%a in ("*.mts") do ffmpeg.exe -i "%%a" -vcodec libxvid -s 848x480 -maxrate 4800k -b 4000k -qmin 3 -qmax 5 -bufsize 4096 -acodec libmp3lame -ac 2 -ab 160k -y "%%~na_xvid.avi"

I have very little background in image editing - I found the above script through some google / forum searching, and it served my purposes quite well.

However, for the paper which I am writing, I have been asked to mention what degree of loss there was from the conversion, and whether that had any impact on my results. I know from the end result that there was no impact, however, I would still like to be able to mention in the paper a more accurate response as to what the percentage of loss was from the conversion.

Can anyone help me out with this? Thanks in advance!

  • Audio does not concern me for this paper. Regarding why MPEG-4 Part 2, the MATLAB script which I had written required the video to be in AVI format. Also, the video conversion is not being done for human viewing - I have a MATLAB script which then does a pixel-by-pixel analysis of the video. Regarding the spatial loss, what sort of loss are we referring to? Edges being cut off? – Kiran Ramaswamy Dec 1 '14 at 15:09
1

Your paper is asking for an impossible data point unless you have some method of evaluation specified. By definition, lossy compression results in a loss of data. Modern lossy compression is very good at discarding data that is not visually meaningful to people, but it is still discarding data.

The amount of data loss when related to a human viewer is thus very subjective and can't be measured. There are specific methods for comparing the source and output videos for differences and determining loss mathematically, but you need a particular choice of method to apply and you simply apply it.

  • I'm not sure what you mean by method of evaluation. To clarify what I'm using this conversion for, my original video was recorded using a JVC camcorder in .MTS, and the converted video was read into MATLAB for use in a pixel-by-pixel analysis algorithm. – Kiran Ramaswamy Dec 1 '14 at 15:10
  • @KiranRamaswamy then perhaps they want you to come up with your own approach for evaluating the difference. When you use lossy encoding, the input and output may look similar, but at a pixel by pixel level, they will be vastly different due to quantization error (which is why the compression works). – AJ Henderson Dec 1 '14 at 16:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.