Chromatic aberration? at DVinfo.net

David Deehan · July 27th, 2004, 01:13 PM

Could someone explain how I could minimise green fringing? This mostly seems to occur at the body edges of subjects (with tanned skin) shot against a contrasting background (e.g. blue sky or water). It is not evident on subjects with pale skin colour. The links below demonstrate this.

http://homepage.ntlworld.com/david.d...ics/Green1.jpg (edges of forearms)
http://homepage.ntlworld.com/david.d...ics/Green2.jpg (shoulders, arm and back)[/url]
http://homepage.ntlworld.com/david.d...ics/Green3.jpg (shoulders and arms)

Is this due to chromatic aberration, poor white balance adjustment or something completely different?

David

Andre De Clercq · July 27th, 2004, 02:36 PM

For shure there is no chromatic aberration involved. Are these pics original DV footage downloaded through firewire?..I doubt. What you get is chroma shift. (mainly vertical shift). Although Pal has an intrinsic vertical chroma shift when decoded, the magnitude of the shift looks as if those pics are not original DV footage but transported via analog video. The only way to correct this is by using a TBC with chroma shift features

David Deehan · July 27th, 2004, 02:56 PM

Hello again Andre,

Yes, the pics are all captured from original DV footage from a Canon XM2 downloaded through firewire.

I don’t know if you remember, but you replied to me in a previous thread (http://www.dvinfo.net/conf/showthread.php?s=&threadid=7866). The topic then referred to “red colour bleed” on a Canon XM2. You think this green fringe artefact is another example of the same problem. I didn’t suspect that (Oh dear!). Following your last advice I sent the camera to Canon for inspection. Their reply was that “they could find no fault and that the camera was to factory specification”.

regards,

David

Andre De Clercq · July 27th, 2004, 03:46 PM

This is indeed a kind of bleed too. A pure bleed should not contain chroma shifts However here you have the aditional shift which make the bleed effects more visible. Also veryfy yr NLE performance. DV codecs used in earlier NLE's sometimes generated (slight) shifts and bleeding too. Do you see the same effects when playing back staight from the camcorder (through analog) to yr TV?

David Deehan · July 27th, 2004, 04:17 PM

Andre,

Yes, the green fringe is also visible when the camcorder is footage is played back straight to a TV, although less visible due to the TV’s reduced resolution. On a high-resolution Trinitron computer screen it is very noticeable.

In my previous thread, the red shift described was always to the left and bottom edges of red objects. This green shift seems to protrude to the top and right, but only seems to originate from tanned (brown) skin colour. Due to the latter I didn’t suspect a Chroma Shift problem.

David

Andre De Clercq · July 28th, 2004, 03:39 AM

David, I searched yr previous posting, and the images you got now are in perfect simularity with what you get today: excessive chroma shift, (mainly down and slightly left) on the chroma vs the luma component. Remember that tanned skin is (mainly) being composed out of red and green (in fact R-Y and B-Y). If the red is shifted down you only get the green component in the corresponsing transition area. Although it is not difficult to further quantify this problem (on pics shooted from color testpatterns), I would advise you to find a simular cam and visually compare pics. For shure my VX2K (Sony) doesn't show these shifts.

David Deehan · July 28th, 2004, 03:07 PM

Andre,

Thanks very much for spending some your time investigating this problem. Taking into consideration comments from other people and my own experience it would seem that chroma shift is an intrinsic problem with the Canon XM2. Canon Service did not seem interested and brushed me off with a generic response. The camera is now well out of warranty so it looks like I’ll just have to put up with this defect until my next purchase (which will probably be a Sony).

David

Cosmin Rotaru · August 6th, 2004, 03:38 AM

David, I'm not sure your camcorder has a problem :) I think all camcorders have this issue. Mine does. It's just that 99% of the time you don't notice it. It's just when you have a particular image like yours (or as I tried - a face on white background) that it's so obviouse and annoying...
It has to be a software solution for this. I'll have to look into this. It will probably take some rendering time, but I would just do it on those particular images I've talked about.

July 27th, 2004, 01:13 PM	#1
David Deehan Regular Crew Join Date: Feb 2003 Location: Belfast (Northern Ireland) Posts: 26	Chromatic aberration? Could someone explain how I could minimise green fringing? This mostly seems to occur at the body edges of subjects (with tanned skin) shot against a contrasting background (e.g. blue sky or water). It is not evident on subjects with pale skin colour. The links below demonstrate this. http://homepage.ntlworld.com/david.d...ics/Green1.jpg (edges of forearms) http://homepage.ntlworld.com/david.d...ics/Green2.jpg (shoulders, arm and back)[/url] http://homepage.ntlworld.com/david.d...ics/Green3.jpg (shoulders and arms) Is this due to chromatic aberration, poor white balance adjustment or something completely different? David

August 6th, 2004, 03:38 AM	#8
Cosmin Rotaru Major Player Join Date: Feb 2003 Location: Romania, Timisoara Posts: 453	David, I'm not sure your camcorder has a problem :) I think all camcorders have this issue. Mine does. It's just that 99% of the time you don't notice it. It's just when you have a particular image like yours (or as I tried - a face on white background) that it's so obviouse and annoying... It has to be a software solution for this. I'll have to look into this. It will probably take some rendering time, but I would just do it on those particular images I've talked about. __________________ Cosmin Rotaru

July 27th, 2004, 02:36 PM	#2
Andre De Clercq Major Player Join Date: Sep 2002 Location: Belgium Posts: 804	For shure there is no chromatic aberration involved. Are these pics original DV footage downloaded through firewire?..I doubt. What you get is chroma shift. (mainly vertical shift). Although Pal has an intrinsic vertical chroma shift when decoded, the magnitude of the shift looks as if those pics are not original DV footage but transported via analog video. The only way to correct this is by using a TBC with chroma shift features

July 27th, 2004, 02:56 PM	#3
David Deehan Regular Crew Join Date: Feb 2003 Location: Belfast (Northern Ireland) Posts: 26	Hello again Andre, Yes, the pics are all captured from original DV footage from a Canon XM2 downloaded through firewire. I don’t know if you remember, but you replied to me in a previous thread (http://www.dvinfo.net/conf/showthread.php?s=&threadid=7866). The topic then referred to “red colour bleed” on a Canon XM2. You think this green fringe artefact is another example of the same problem. I didn’t suspect that (Oh dear!). Following your last advice I sent the camera to Canon for inspection. Their reply was that “they could find no fault and that the camera was to factory specification”. regards, David

July 27th, 2004, 03:46 PM	#4
Andre De Clercq Major Player Join Date: Sep 2002 Location: Belgium Posts: 804	This is indeed a kind of bleed too. A pure bleed should not contain chroma shifts However here you have the aditional shift which make the bleed effects more visible. Also veryfy yr NLE performance. DV codecs used in earlier NLE's sometimes generated (slight) shifts and bleeding too. Do you see the same effects when playing back staight from the camcorder (through analog) to yr TV?

July 27th, 2004, 04:17 PM	#5
David Deehan Regular Crew Join Date: Feb 2003 Location: Belfast (Northern Ireland) Posts: 26	Andre, Yes, the green fringe is also visible when the camcorder is footage is played back straight to a TV, although less visible due to the TV’s reduced resolution. On a high-resolution Trinitron computer screen it is very noticeable. In my previous thread, the red shift described was always to the left and bottom edges of red objects. This green shift seems to protrude to the top and right, but only seems to originate from tanned (brown) skin colour. Due to the latter I didn’t suspect a Chroma Shift problem. David

July 28th, 2004, 03:39 AM	#6
Andre De Clercq Major Player Join Date: Sep 2002 Location: Belgium Posts: 804	David, I searched yr previous posting, and the images you got now are in perfect simularity with what you get today: excessive chroma shift, (mainly down and slightly left) on the chroma vs the luma component. Remember that tanned skin is (mainly) being composed out of red and green (in fact R-Y and B-Y). If the red is shifted down you only get the green component in the corresponsing transition area. Although it is not difficult to further quantify this problem (on pics shooted from color testpatterns), I would advise you to find a simular cam and visually compare pics. For shure my VX2K (Sony) doesn't show these shifts.

July 28th, 2004, 03:07 PM	#7
David Deehan Regular Crew Join Date: Feb 2003 Location: Belfast (Northern Ireland) Posts: 26	Andre, Thanks very much for spending some your time investigating this problem. Taking into consideration comments from other people and my own experience it would seem that chroma shift is an intrinsic problem with the Canon XM2. Canon Service did not seem interested and brushed me off with a generic response. The camera is now well out of warranty so it looks like I’ll just have to put up with this defect until my next purchase (which will probably be a Sony). David