Purple fringing and FX1

Patrick Swinnea · May 4th, 2005, 03:12 AM

Great work Andrew! Very well done. I really liked the test at different focal lengths. It looks like if you hang towards the middle (and avoid shooting into the sun) you should be fine. On projects intended for film-outs this is a very easy workaround for a spectacular image. For projects not intended for a huge blow-up, I doubt anyone would even see it.

Also, I've started looking specifically for CA in downloaded clips and realized you really have to be looking to find it, and most shots don't show it at all.

There are two FX1/Z1 samples from Germany about half the way down this page -

http://www.highdefforum.com/showthread.php?t=1310

The CA is there in *maybe* three shots, which proves that 1) It IS FX1/Z1 footage and 2) I should really stop worrying about something this small. This camera can make beautiful pictures. I'm sold.

I look forward to your focus / resolutions tests. Thanks again.

Michael Struthers · May 4th, 2005, 10:46 AM

Sony's purple fringing issue is much more important for still photographers - if you study a photo long enough you can find it - if it's there.

However, I defy a consumer to find "purple fringing" while watching motion video, you are noticing the action (or lack thereof) in the frame.

Augusto Manuel · May 5th, 2005, 12:28 AM

I have been in this business primarily broadcast for quite a while.

Could anyone really explain to me why people pay so much attention to pixel count of the CCD when it comes to small digital cameras versus the actual lines of horizontal / vertical resolution of the camera which broadcast cameras are really measured when it comes to resolution ????

Since we have all become system analysts instead of TV engineers. Remember that the TV raster in NTSC is measured in terms of lines of resolution not pixel count like a computer monitor.

Also remember that CCD pixel count sometimes is quite irrelevant because not all pixels are counted when it comes to resolution.

So I would really like to know from someone who has some TV engineering background .

Quote:

Originally Posted by Tim Ashbrooke

Hi guys

I have an FX1E and noticed that on some high contrast edges in HDV mode, there's purple fringing. This is a phenomenon of high pixel count digital cameras. I am assuming that this is happening because of the high pixel count of the FX1's CCDs. Anyone had problems with this phenomenon?

cheers
Tim

Jimmy McKenzie · May 5th, 2005, 07:15 AM

I suppose this is a by-product of the media convergence that has seen us become adept with computer programs that transform our images to print from vector, print from raster, 3d modeling and motion graphics all in an effort to support our finished stream of pixel originated materials.
Sure the finished mpeg on my dvd is made to display on an ntsc twin field tube, but everything that occurs prior to that in the computer is completely based upon codecs and raster imagery that is pixel based for our manipulation. Since this multimedia convergence always includes more than just a dv stream, I find it extremely important to know how certain products affect my work. The purple fringing phenomenon is a mammoth blight when one is working with certain brands of digital still cameras in the prosumer price range that simply don't measure up to an acceptable standard.

Still photos are a big part of what I do. So to answer your question, the standard way of describing the digital merits of our gear has found it's common denominator. The ubiquitous pixel.

Hmmmm. Perhaps the next dv indie challenge could use "the ubiquitous pixel" as the next challenge keyword.

Patrick Swinnea · May 5th, 2005, 08:11 AM

I don't think Tim's post was meant to really get into pixel count. I think he was just making a note that this is a problem with some high pixel count cameras, and was throwing it out there as a possible solution to the "purple fringe mystery."

If this camera shot 320 x 240 and had purple fringe, we'd be having the same conversation, IMO.

Augusto Manuel · May 5th, 2005, 10:53 AM

I really don't think pixel count has that much importance for video cameras. Since my old days we have measured camera resolution in lines and that's the way TV engineers still measure these days. CCD pixel count is not so relevant when measuring resolution because you have technologies such pixel offset, and many other schemes in pixel extrapolation which makes pixel count in a way irrelevant. Sure pixel rich cameras such as F900s have more pixels than Z1 camera imagers and more resolution, but that does not tell me exactly where resolution wise they both stand.

I think it is another area such as lux rating which is irrelevant when measuring the minimum illumination of a camera for comparison purposes.

And I agree with you about that this is a by-product of the media convergence. However, I would tend to think it is the wrong way to measure resolution of cameras for comparison purposes. Camera CCDs and TV rasters are not computer monitors. The NTSC equipment uses lines of resolution instead of pixel resolution. I know many programs go for the web these days instead of broadcast or playback thru a deck, but cameras are still made within the NTSC system and need to be judged as such. Have you seen a manual saying the resolution of this camera is 800,000 pixels. I don't remember. I have seen, however, camera manuals displaying the line resolution of the equipment and that's has been fairly common.

Quote:

Originally Posted by Jimmy McKenzie

I suppose this is a by-product of the media convergence that has seen us become adept with computer programs that transform our images to print from vector, print from raster, 3d modeling and motion graphics all in an effort to support our finished stream of pixel originated materials.
Sure the finished mpeg on my dvd is made to display on an ntsc twin field tube, but everything that occurs prior to that in the computer is completely based upon codecs and raster imagery that is pixel based for our manipulation. Since this multimedia convergence always includes more than just a dv stream, I find it extremely important to know how certain products affect my work. The purple fringing phenomenon is a mammoth blight when one is working with certain brands of digital still cameras in the prosumer price range that simply don't measure up to an acceptable standard.

Still photos are a big part of what I do. So to answer your question, the standard way of describing the digital merits of our gear has found it's common denominator. The ubiquitous pixel.

Hmmmm. Perhaps the next dv indie challenge could use "the ubiquitous pixel" as the next challenge keyword.

Jimmy McKenzie · May 5th, 2005, 12:03 PM

I agree.
What will become irrelevant or obsolete is the interlaced line.
Progressive video HDTV will see to that.
Then, the manufacturers will have to pay very close attention to their pixels and quality of their imaging glass and chips.
The example noted in this thread is the first of many comparisons that will separate the good, bad and ugly no matter what the nomenclature.
That's my perception.
I have 0 experience with videotape editing. I couldn't pick out an A/B roll machine if were in a room filled with cabbages. For me, the old days were when powerpoint was displayed through a projector and somebody asked what if that were made to also play on a tv. It wasn't dazzling. Pun intended.
So far, I have been very hesitant to move from SD to HD since many of the unknown knowns aren't really known yet. What I do know is that I won't repeat the CA gaff that caught me off guard when I unknowingly upgraded to 8 megapixels for my single frame non-interlaced captures.

Thomas Smet · May 5th, 2005, 05:44 PM

I always thought this was a by-product of pixel shift because the green channel is offset by half a pixel so you get a half a pixel purple edge on certain colors.

When Juan was first working on his DVX-100A adapter for uncompressed video we played around with still images from the early adapter to try and reverse the pixel shift. When the pixel shift was removed most of the purple fringing was gone.

The main difference between the DVX-100A and the Z1 is the fact that the DVX-100A mainly used pixel shift to sharpen the image. The CCD's already had enough resolution for a great image. The Z1 on the other hand is using pixel shift to gain raw resolution.

I could be wrong about this but I did notice the purple fringing to go away when pixel shift was removed.

Andrew J Hall · May 5th, 2005, 06:41 PM

I think the fact that you can see green / purple fringing pretty clearly indicates that this is chromatic aberration though perhaps it is exacerbated by the pixel shifting technique - which I do not yet understand (anyone got a good reference).

Andrew

John Beale · May 5th, 2005, 07:03 PM

I did some shots today to evaluate the Century Optics 0.6x wide adaptor (made for the DVX100) for use on the FX1. It seems to work OK, adding some barrel distortion of course.

However, in the "without wide adaptor" comparison shot I took, I can clearly see purple and green fringing where the sky is visible through the gaps in a fence, towards the edges of the frame. It is not visible in the center of the frame although the sky is just as bright there. There is a purple fringe on one side of the gap and green on the other, on the left side of the frame, with the colors reversed on the other side, and no fringe in the middle. Based on these observations I attribute this to chromatic abberation in the lens. You can see the test shot here:

http://bealecorner.net/D30/050505/03-norm.htm

Dan Diaconu · May 5th, 2005, 09:12 PM

I just have noticed the above by accident during some other tests.
In this pic:
http://dandiaconu.com/gallery/album05/IMGA0412?full=1
look for two square sections at the bottom of the mirror.
Left one has some minor color fringe and the right one is even worse. Same lens, same all....different contrast. Also on the rest of the image, wherever the contrast exceded the ability (for that particular speed/aperture setting) of the ccd to respond, the color fringe is there. Zebra helps.
Even the pic from the previous post reinforces (IMHO)this "confusion" (conclusion?)

Jimmy McKenzie · May 6th, 2005, 03:51 AM

Dan you have uncovered the hidden feature that is very similar to the others, apart from the usually accompanying green fringe. Whether or not this is lens or chip or otherwise electronically created, if you call the manufacturer and describe the problem they will tell you a) it's normal and b) avoid shooting areas of high contrast and to avoid low f-stop numbers.
For convenience sake, they call it CA.
I am curious with regard to the size of the frame grab. Where does 1720x1360 come from?

Dan Diaconu · May 6th, 2005, 09:51 AM

Jimmy, I never thought about calling them. Their answer is right (imo) and any lens behavior is the same (more on the edges). When the "feature" is in the center but under proper lighting conditions disappears, the lens is clearly not the offender... Frame size is a grab from GS200's still camera (2.3Mpix)
On the same theme, the rez chart would be better if printed on gray BG instead of white (imo) That would lower the contrast, bringing it in the CCD's ability to respond without blooming and making it more fair to judging the lens ability to resolve fine lines (if lens is questioned) rather than....oh well... about 600 lines...(just because the CCD is blooming and creating artificial edges) When the lines reach the one pixel size, the resolution is half (imo) to what the lens can deliver. Why? Because the other half is lost in the blooming of the neighboring area.
(you could see a slight resolution increase if you dim the light while shooting a rez chart)

May 4th, 2005, 03:12 AM	#16
Patrick Swinnea Regular Crew Join Date: Apr 2004 Location: Denver, CO, USA Posts: 93	Great work Andrew! Very well done. I really liked the test at different focal lengths. It looks like if you hang towards the middle (and avoid shooting into the sun) you should be fine. On projects intended for film-outs this is a very easy workaround for a spectacular image. For projects not intended for a huge blow-up, I doubt anyone would even see it. Also, I've started looking specifically for CA in downloaded clips and realized you really have to be looking to find it, and most shots don't show it at all. There are two FX1/Z1 samples from Germany about half the way down this page - http://www.highdefforum.com/showthread.php?t=1310 The CA is there in maybe three shots, which proves that 1) It IS FX1/Z1 footage and 2) I should really stop worrying about something this small. This camera can make beautiful pictures. I'm sold. I look forward to your focus / resolutions tests. Thanks again. Last edited by Patrick Swinnea; May 4th, 2005 at 07:53 AM.

May 4th, 2005, 10:46 AM	#17
Michael Struthers Major Player Join Date: Jul 2003 Location: SF, Ca Posts: 421	Sony's purple fringing issue is much more important for still photographers - if you study a photo long enough you can find it - if it's there. However, I defy a consumer to find "purple fringing" while watching motion video, you are noticing the action (or lack thereof) in the frame. __________________ Michael Struthers www.buzzdigital.com

May 5th, 2005, 06:41 PM	#24
Andrew J Hall Regular Crew Join Date: Feb 2005 Location: Wellington, New Zealand Posts: 80	I think the fact that you can see green / purple fringing pretty clearly indicates that this is chromatic aberration though perhaps it is exacerbated by the pixel shifting technique - which I do not yet understand (anyone got a good reference). Andrew __________________ http://www.ps-scripts.com/ Intel Core2 Quad Q6600 2.4GHz, 2x 2Gb Ram, ATI Radeon HD 3850

May 5th, 2005, 07:03 PM	#25
John Beale Regular Crew Join Date: Dec 2002 Location: San Francisco Posts: 74	purple/green fringes at edges of image I did some shots today to evaluate the Century Optics 0.6x wide adaptor (made for the DVX100) for use on the FX1. It seems to work OK, adding some barrel distortion of course. However, in the "without wide adaptor" comparison shot I took, I can clearly see purple and green fringing where the sky is visible through the gaps in a fence, towards the edges of the frame. It is not visible in the center of the frame although the sky is just as bright there. There is a purple fringe on one side of the gap and green on the other, on the left side of the frame, with the colors reversed on the other side, and no fringe in the middle. Based on these observations I attribute this to chromatic abberation in the lens. You can see the test shot here: http://bealecorner.net/D30/050505/03-norm.htm

May 5th, 2005, 09:12 PM	#26
Dan Diaconu Major Player Join Date: Dec 2004 Location: Vancouver, Canada Posts: 804	Color fringe seems to be hi contrast image related rather than lens aberrations I just have noticed the above by accident during some other tests. In this pic: http://dandiaconu.com/gallery/album05/IMGA0412?full=1 look for two square sections at the bottom of the mirror. Left one has some minor color fringe and the right one is even worse. Same lens, same all....different contrast. Also on the rest of the image, wherever the contrast exceded the ability (for that particular speed/aperture setting) of the ccd to respond, the color fringe is there. Zebra helps. Even the pic from the previous post reinforces (IMHO)this "confusion" (conclusion?)

May 5th, 2005, 07:15 AM	#19
Jimmy McKenzie Major Player Join Date: Jan 2004 Location: Waterloo Ontario Posts: 721	I suppose this is a by-product of the media convergence that has seen us become adept with computer programs that transform our images to print from vector, print from raster, 3d modeling and motion graphics all in an effort to support our finished stream of pixel originated materials. Sure the finished mpeg on my dvd is made to display on an ntsc twin field tube, but everything that occurs prior to that in the computer is completely based upon codecs and raster imagery that is pixel based for our manipulation. Since this multimedia convergence always includes more than just a dv stream, I find it extremely important to know how certain products affect my work. The purple fringing phenomenon is a mammoth blight when one is working with certain brands of digital still cameras in the prosumer price range that simply don't measure up to an acceptable standard. Still photos are a big part of what I do. So to answer your question, the standard way of describing the digital merits of our gear has found it's common denominator. The ubiquitous pixel. Hmmmm. Perhaps the next dv indie challenge could use "the ubiquitous pixel" as the next challenge keyword.

May 5th, 2005, 08:11 AM	#20
Patrick Swinnea Regular Crew Join Date: Apr 2004 Location: Denver, CO, USA Posts: 93	I don't think Tim's post was meant to really get into pixel count. I think he was just making a note that this is a problem with some high pixel count cameras, and was throwing it out there as a possible solution to the "purple fringe mystery." If this camera shot 320 x 240 and had purple fringe, we'd be having the same conversation, IMO.

May 5th, 2005, 12:03 PM	#22
Jimmy McKenzie Major Player Join Date: Jan 2004 Location: Waterloo Ontario Posts: 721	I agree. What will become irrelevant or obsolete is the interlaced line. Progressive video HDTV will see to that. Then, the manufacturers will have to pay very close attention to their pixels and quality of their imaging glass and chips. The example noted in this thread is the first of many comparisons that will separate the good, bad and ugly no matter what the nomenclature. That's my perception. I have 0 experience with videotape editing. I couldn't pick out an A/B roll machine if were in a room filled with cabbages. For me, the old days were when powerpoint was displayed through a projector and somebody asked what if that were made to also play on a tv. It wasn't dazzling. Pun intended. So far, I have been very hesitant to move from SD to HD since many of the unknown knowns aren't really known yet. What I do know is that I won't repeat the CA gaff that caught me off guard when I unknowingly upgraded to 8 megapixels for my single frame non-interlaced captures.

May 5th, 2005, 05:44 PM	#23
Thomas Smet Trustee Join Date: Mar 2004 Location: Milwaukee, WI Posts: 1,719	I always thought this was a by-product of pixel shift because the green channel is offset by half a pixel so you get a half a pixel purple edge on certain colors. When Juan was first working on his DVX-100A adapter for uncompressed video we played around with still images from the early adapter to try and reverse the pixel shift. When the pixel shift was removed most of the purple fringing was gone. The main difference between the DVX-100A and the Z1 is the fact that the DVX-100A mainly used pixel shift to sharpen the image. The CCD's already had enough resolution for a great image. The Z1 on the other hand is using pixel shift to gain raw resolution. I could be wrong about this but I did notice the purple fringing to go away when pixel shift was removed.

May 6th, 2005, 03:51 AM	#27
Jimmy McKenzie Major Player Join Date: Jan 2004 Location: Waterloo Ontario Posts: 721	Dan you have uncovered the hidden feature that is very similar to the others, apart from the usually accompanying green fringe. Whether or not this is lens or chip or otherwise electronically created, if you call the manufacturer and describe the problem they will tell you a) it's normal and b) avoid shooting areas of high contrast and to avoid low f-stop numbers. For convenience sake, they call it CA. I am curious with regard to the size of the frame grab. Where does 1720x1360 come from?

May 6th, 2005, 09:51 AM	#28
Dan Diaconu Major Player Join Date: Dec 2004 Location: Vancouver, Canada Posts: 804	Jimmy, I never thought about calling them. Their answer is right (imo) and any lens behavior is the same (more on the edges). When the "feature" is in the center but under proper lighting conditions disappears, the lens is clearly not the offender... Frame size is a grab from GS200's still camera (2.3Mpix) On the same theme, the rez chart would be better if printed on gray BG instead of white (imo) That would lower the contrast, bringing it in the CCD's ability to respond without blooming and making it more fair to judging the lens ability to resolve fine lines (if lens is questioned) rather than....oh well... about 600 lines...(just because the CCD is blooming and creating artificial edges) When the lines reach the one pixel size, the resolution is half (imo) to what the lens can deliver. Why? Because the other half is lost in the blooming of the neighboring area. (you could see a slight resolution increase if you dim the light while shooting a rez chart)