?ChromaInterlace effect

[WorBry]

Hi,

I wonder if someone could advise me. I'm capturing PAL miniDV home videos from my Sony DCR-PC115E digital camcorder using an ADS Pyrolink firewire card and either WideoWave or Pinnacle Sudio. Both the original captures and DV edits tend have a greenish 'bleed' on the upper edge of pale objects such as skin, regardless of whether I interconvert between Type I and Type II. It also persists, although marginally less so, when I convert to interlaced DVD (M2V+PCM) using TMPGenc. More recently I've been converting the Type II files to DivX at relatively high bitates (2200 kbps, 2-pass with B-frames)using AVISynth scripts and VDubMod. I've experimented with various deinterlacers and concluded that KernelDeint gives the sharpest image overall. However, the green margin persists. I've checked the field order (bottom first) to ensure the right deinterlace settings (Order=0,sharp=true) and am resizing correctly with Lanczos. Pre-conversion to YUY2 or YV12 colorspace seems to make no difference. Interestingly, the green-edge effect is much reduced with FieldDeinterlace and disappears if I first SeparateFields (i.e a 50fps progressive encode). Since FieldDeinterlace includes a ChromaDeinterlace function, this makes me wonder if this is some sort of chroma-interlace effect. I dont really want to change to FieldDeinterlace as I find the clarity of KernelDeint generally superior. Can anyone suggest a workaround for this? Cheers.

[MGRip]

Hi,
Can you upload an image of greenish bleeding?

If I understood you well, I think I know what is the problem.
The problem is within your camcorder, your pc115, mine pc5 etc. are single CCD camocrders, wich is more difficult to guess the true color. Single CCD captures the video as black and white, then, guesses the color for each channel. 3CCD camcorder works different, it has a crystal that divides the light into 3 channels to 3 different CCDs, each uses his luma information (0-255) to create combined true color, it's like 3 camcorders in one.

1CCD camcoder artifacts:
Color distorion (very noticable in low light situations).
Lake of color accuracy (to real colors).

Regards,
Mark.

[WorBry]

Thanks for your reply MGRIP.

I'm not too sure how to upload an image. Anyhow, having read more about the subject, the aberration I'm seeing does appear to be consistent with "color bleeding" to which, as you explained, single CCD camcorders are more or less prone. It's interesting though that the effect is much reduced by FieldDeinterlace and when the fields are separated completely it appears to be eliminated. However, to play the latter 50fps DivX encodes smoothly at the relatively high bitrates I use means a trade-off of quality for performance in playback and they wont play at all on my DivX-DVD standalone.

I'm just wondering if anyone is aware of a post-processing filter or a capture software or method that can counter the "color bleed" effect as I still much prefer KernelDeint for deinterlacing. I've tried playing around with the ChromaShift AVISyth filter but without any success.

[WorBry]

Well, I appear to have resolved the problem by installing the Sony DV Software Codec. Cant see any of the greenish color bleed now in the PAL DV source and DivX encodes. A resounding success!! No more messing around with chroma filters to try and fix it. I can only assume that the problem was associated with the Adaptec DVSoft codec installed with MGI VideoWave. I didnt even have to uninstall it. The Sony codec automatically came up as the default. Thanks for the info about the CCD aspect though.

[MGRip]

WOW, it's very intresting!
Can you upload an image?
(When you reply you have above the submit button an attach file option, click browse, and select an image from your computer)

[WorBry]

I'm afraid I cant put up a frameshot of the source Pal TypeII DV using the Adaptec DVSoft codec because I've removed the codec, but attached are samples of frame shots (JPEG sample from VirtualDub tga frameshot)from DivX videos encoded in VirtualDub; the same path was used for both:

AVS Script:
avisource("E:\test file.avi")
ConvertToYUY2(interlaced=true)
KernelDeint(order=0,sharp=true)
LanczosResize(720,528)
Undot()
VirtualDub:
Fast Recompress
DivX 5.2.1, standard 2-pass, single B-frames, no GMC or QPel

The first image attached here used the Adaptec DVSoft codec for decompression and you should be able to see the greenish 'bleed' on the top edge of the skin. The second image on my following reply used the Sony DV Software codec and there is no bleed.

[WorBry]

Heres the second image i.e. using the Sony DV Software codec. I hope these images come through OK because I've never done this before.

Edit: Cant see the image attachment - help !!

[MGRip]

Maybe the file doesn't have valid extention?
Valid file extensions: gif jpg png txt zip bmp jpeg vcf
And also the maximum size is: 204800 bytes (200 KB).

[WorBry]

I'm not sure what I'm doing wrong. The images were jpg - only 50KB. I've tried uploading them again - just get a message saying they are already loaded. Where are the attachments supposed to appear? Tried converting to bmp but then they were too large - 250KB. I'll have another go tomorrow.

Edit: Tried again. Still cant see any attachments. So you might have to take my word about the "color bleed" aspect. One thing I have noticed about the Sony DV codec though, is that the image has a very slight "mistiness" about it, at least when compared to Adaptec DVSoft - but I think I can live with that as the overall image quality is superior.

Edit 7th Nov. I see the image attachments have now appeared above. That's one of my newborn twin boys.....hence the reinspired interest in home video editing.

[WorBry]

OK, now I'm totally confused.

I recently changed to Ulead VideoStudio 8 as my main video editor (so I could do fancy video overlays !). Captured fresh footage from my PAL DCR-PC115E to PAL Type II using the Ulead capture function. Edited with some text and transitions. "Smart-Rendered" the project as a PAL DV Type II selecting the Sony Software Codec (Type II) and was surprised to note that the edited portions (notably where I had inserted text) were darker than the "untouched" video. So I set the output to total recompress and rendered again with the Sony codec. Low and behold, when loaded in VDub (or played back in CoreMediaPLayer), the "greenish bleed" effect I illustrated above was back - not what I had expected ! So I rendered the project again selecting the original default "DV Encoder Type II", and this time the output DV showed no color instensity changes upon Smart Rendering and no color bleed after a full recompress. However, VDub Information showed the default decompressor for these encodes to be the Sony Software Codec and they exhibited the same slight "mistiness" that I had described previously with the Sony codec as decompressor. To try and figure out what was going on I loaded the encodes into VDub with an AVISouce.Info script. This confirmed that both the "Sony Codec" and "DV Encoder" renderings were RGB24. For good measure I also loaded the captures into Pinnacle Studio 8 to see what would come out. Made a few edits (transitions, text etc) and outputted to DV. In this case, selecting "force settings same as project" defaulted to "DV Encoder". However, if I made no edits to the loaded clip, the output settings defaulted to "WMVideo Advanced Encoder DMO" (an option that was not available in Ulead Video Studio 8)). Regardless, both renderings were free from the "green bleed". However, with the Sony Software DV Codec selected - you've guessed it - "green bleed".

Does anybody know what this "DV Encoder" (Type I and Type II) is that both Ulead and Pinncle default to, and why there should be this strange paradox with the Sony Software DV Codec i.e. it encodes DV files with "green bleed" effect, yet eliminates the effect when used as a decompressor for DV files encoded with Adaptec DVSoft codec. Any ideas also on how to overcome the "misty" appearance of DV files decompressed with the Sony codec, which I assume to be some sort of lumina phenomenon.