bb
12th December 2002, 09:03
Hi folks,
here are some interesting news from my most recent capture tests:
I captured a small clip (just one minute) using the Huffyuv codec:
- WinTV PCI
- Hauppauge's VfW drivers
- RGB colorspace
- 720x576
Then I filtered it using Telecide to get a progressive source, which I saved through Huffyuv again. So far everything's lossless, and you have the typical noise in the video which makes encoding harder.
Then I loaded this progressive Huffyuv file, and saved it
a) using PicVideo MJPEG at quality 19
b) using Panasonic DV codec
The file sizes from a) and b) didn't differ by much, but of course the Huffyuv file was much bigger. Comparing the three files regarding the noise, the DV file turned out to be the best, then Huffyuv, then MJPEG!
That was a surprise, so I decided to DivX encode the three files:
- DivX 5.0.2
- no b-frames, no q-pel, no GMC
- 1-pass at quantizer 2
- no pre-processing, no resizing, no psycho, etc.
As expected, the file from the MJPEG source turned out to be bigger than the Huffyuv original, because the JPEG compression is sensitive to noise. Thus it was bigger than the DivX file which was encoded from the Huffyuv source. The big surprise was the file created from the DV source: it was even smaller than the one created from Huffyuv source. This sounds a little strange to me, because DV uses a compression being similar to JPEG to a certain extent (DCT etc.).
Conclusion: regarding capture quality it seems that we get the following ranking (1 best, 3 worst):
1. DV
2. Huffyuv
3. MJPEG
I will further investigate on this and report my results. I may very well switch from MJPEG to DV for TV captures, if there are no other quirks I don't see yet.
But be aware of the following constraints when using DV for captures:
- You have to capture in RGB
- Resolution is fixed at 720x576 (for PAL, NTSC is 720x480)
- CPU demands are higher than for MJPEG
Any comments?
bb
here are some interesting news from my most recent capture tests:
I captured a small clip (just one minute) using the Huffyuv codec:
- WinTV PCI
- Hauppauge's VfW drivers
- RGB colorspace
- 720x576
Then I filtered it using Telecide to get a progressive source, which I saved through Huffyuv again. So far everything's lossless, and you have the typical noise in the video which makes encoding harder.
Then I loaded this progressive Huffyuv file, and saved it
a) using PicVideo MJPEG at quality 19
b) using Panasonic DV codec
The file sizes from a) and b) didn't differ by much, but of course the Huffyuv file was much bigger. Comparing the three files regarding the noise, the DV file turned out to be the best, then Huffyuv, then MJPEG!
That was a surprise, so I decided to DivX encode the three files:
- DivX 5.0.2
- no b-frames, no q-pel, no GMC
- 1-pass at quantizer 2
- no pre-processing, no resizing, no psycho, etc.
As expected, the file from the MJPEG source turned out to be bigger than the Huffyuv original, because the JPEG compression is sensitive to noise. Thus it was bigger than the DivX file which was encoded from the Huffyuv source. The big surprise was the file created from the DV source: it was even smaller than the one created from Huffyuv source. This sounds a little strange to me, because DV uses a compression being similar to JPEG to a certain extent (DCT etc.).
Conclusion: regarding capture quality it seems that we get the following ranking (1 best, 3 worst):
1. DV
2. Huffyuv
3. MJPEG
I will further investigate on this and report my results. I may very well switch from MJPEG to DV for TV captures, if there are no other quirks I don't see yet.
But be aware of the following constraints when using DV for captures:
- You have to capture in RGB
- Resolution is fixed at 720x576 (for PAL, NTSC is 720x480)
- CPU demands are higher than for MJPEG
Any comments?
bb