View Full Version : Colour bleed


David Deehan
March 19th, 2003, 05:17 PM
Hi,

This is the first time I have posted to this forum and being a novice with DV technology I do so with a little trepidation.

Last September I purchased a Sony TRV25. I was extremely impressed with this little camcorder however, in December I decided to take the plunge and upgrade further to an XM2. Anyway, my concern is as follows. After two months of testing the XM2 and comparing its results with those of the Sony there is one area where the Sony excels, namely lack of colour bleed.

I have posted some captured frames, which I think demonstrate this effect. These can be viewed at
http://homepage.ntlworld.com/david.deehan/XM2/index.htmlBy the way this web page is another first for me so here’s hoping it works OK.

In a recent thread Davide Manni (GL” Red Push Problem) mentioned the site http://www.gl1-411.com/forum/message.asp?MessageID=1473. This seems to reflect the same problem.

Any help would be greatly appreciated.

David

Don Palomaki
March 19th, 2003, 07:01 PM
Can you please tell us how the images were captured and prepared for the web?

Ken Tanaka
March 19th, 2003, 07:29 PM
As Don noted, we really need to know more about (a) how you shot the samples and, more importantly (b) how you prepared these graphics.

I noticed that your samples are JPEG's. One of the unusual collateral aspects of compressing an interlaced image with this method is that the algorithm can "smear" edges where it finds motion comb.

BTW, regarding the message on the other site that you referrred us to, just for the record the poster's allusion to Canon's pixel shift contributing to such an effect is misinformed. "Pixel shift" deals with a diffferent aspect of the image.

Certainly, I am not saying that you aren't seeing what you claim. We just need to know more about where you're really seeing it and how you're using the camera's controls.

Welcome again, David!

Lars Siden
March 20th, 2003, 04:27 AM
Hello David,

I'm only guessing here ( my main knowledge is digital-still pictures, using the Canon D30 ).

Have you gained the color ? ( in the color gain menu ? ) if so, maybe you have gained it too much. Red is a sensitive colour.

Have you changed the sharpness settings on the camera? If you have "softened" the picture, try to sharpen it, or vice versa.

Did you use a program mode on the camera? Or where you on manual settings? Did you use a light-profile ( indoor/outdoor/tungsten etc ) or did you calibrate the whitebalance manually?

Hope these questions can help you try more ways of fixing the colourbleed.

Best regards,

Lars Siden

Stockholm, Sweden

Andre De Clercq
March 20th, 2003, 05:53 AM
Several elements come into play:
-Camera type (pixel shift...)
-Color encoder (NTSC/PAL) quality
-Saturation level
-Y/C delay
-Display decoder (PAL/NTSC) Quality
-DV codec (NTSC 4:1:1) which lowers Hor color res to 25% of luma
(means color leakage)
To verify the original footage one should playback on a DV component player and go straight into a component display in order to avoid/exclude the color encoder/decoder shortcommings

David Deehan
March 20th, 2003, 08:01 AM
Thanks for all the replies.

Here is some detail of the camera settings, video and still capture configuration, etc.

Camera settings:
XM2 - Pal
Camera mode – Normal
Recording program – TV (shutter speed 120)
White balance – Automatic
Optical stabilizer – On
Auto exposure – On (i.e. exposure lock off)
Auto focus – On
White balance – Auto
AE shift – default
Color Gain – default
Sharpness – increased 1 point
Setup Level – decreased 2 points

A UV filter is attached.

Video capture:
Hardware – firewire, Pentium IV – 2GHz, Matrox G400 video card, Iiyama 19” monitor.
Software – Vegas Video, Windows XP.


Still capture:
Frames were captured with Vegas Video clipboard capture facility and pasted into Paint Shop Pro
Original clips were used (no compression or deinterlacing)


Earlier today I played the clip (in windows Media Player) from which Smeared Image 2 was captured (see web page address in original post). I paused this clip and compared the frame side by side with the captured JPEG. The smearing artefacts were identical on both.

I also did the above comparison using a laptop with TFT display and achieved similar results.

I haven’t as yet watched any of these clips on television as its poorer resolution would probably mask much of the bleeding. I’ll probably do that tonight after work.

Thanks again,

David

Lars Siden
March 20th, 2003, 12:16 PM
Hello David,

I see nothing strange in your setup. The only thing that strikes me as possible is that your version of Vegas(or some other Video editing software ) has changed your windows DV drivers. If possible, try to transfer some video to another machine, and see if the bleeding is still there ... if so - get it a plaster *grin*

// Lazze

Tom Christensen
March 21st, 2003, 09:34 AM
I asked the same question when I experienced this on my GL2 and I got the same responses on this board.

http://www.dvinfo.net/conf/showthread.php?s=&threadid=4970&highlight=red+ghosting

Put a piece of red tape on a white background and you'll really get the effect (and probably get more annoyed like I did).


I usually try to turn down the color gain a bit and be aware when shooting anything red/orange.

I've learned to live with it because I really like the camera, but it is annoying since my 1ccd Sony doesn't seem to have this problem.

Tom

Rob Lohman
March 21st, 2003, 10:16 AM
I have some problems with reds on my XL1s as well (striping
and same kind of bleeding as you do).. Seems to be isolated
to the reds..... Funny thing is that whenever I encode it to MPEG2
it seems to be gone. Weird.

David Deehan
March 21st, 2003, 02:08 PM
I see Tom Christensen has mentioned this problem in another thread. If it’s a specific effect restricted to the pixel shift technology used in Canon cams, I’m surprised it hasn’t been highlighted more often. I’m also surprised not seeing it mentioned in any reviews. It may have swayed my buying decision towards a Sony model. I can concur with the end of Tom’s final sentence – “my 1ccd Sony doesn't seem to have this problem”. My previous Sony TRV25 did not show any bleeding effects.

I also agree with Tom’s observation in the thread http://www.dvinfo.net/conf/showthread.php?s=&threadid=4970&highlight=red+ghosting , “I tried the red tape on a white and then black background and Yikes!!!, fringing was very obvious. It became worse the farther away the tape was, so close up it was slight but moving away from the tape, the thing glowed.”

Since my last post I connected the XM2 to a television and observed the “bleeding” artefact. This would rule out computer hardware and software bugs.

Bleeding in the captured frame “Smeared Image 2” on my web site http://homepage.ntlworld.com/david.deehan/XM2/index.html is not restricted to “red”. Both “blue” (sign on wall) and “green” (post-box) also show signs of it.

The frames I have chosen were from clips handpicked to demonstrate the problem. These were extreme examples. Most of my filming to date does not show bleeding to such a degree; but is often there to a lesser degree. It seems to diminish image sharpness. I will turn down the colour gain and see what happens.

Anyway, its Friday night so I’m of to the bar!!!

Thanks for all your comments,

David

Jeff Donald
March 21st, 2003, 03:10 PM
Try the 85B filter it works to lessen the problem in some of the DVX100 cameras. I have a client who swears by using the 85B the fringing is considerably lessened. There are plugins for FCP that may help also (if you use a Mac).

Bud Kuenzli
March 22nd, 2003, 05:36 PM
what plugins are you referring to for FCP and fringing?

Gabriel_Knight
March 23rd, 2003, 04:06 AM
The smearing of colors is very annoying, and I am glad this camera does not have a lot of smearing.

My previous camera (a Sony! TRV20) had a lot of smearing!
It was much worse than the XM2 I now have. When you tried to film a white duck in brown/black water, you will get a lot of smearing with the sony, but almost none with the XM2.

Jeff Donald
March 23rd, 2003, 01:41 PM
Joe's Filters (http://www.joesfilters.com/) has a Y/C filter to help with retiming color.

David Deehan
October 22nd, 2003, 11:20 AM
Hi everyone,

apologies for resurrecting this thread, but I’m about to contact Canon support prior to the expiry of the warranty. Since my last post in March I have updated my website ( http://homepage.ntlworld.com/david..../XM2/index.html ) with further images demonstrating colour bleed – namely “Smeared Image 3” and 6 new image links at the bottom of the page. I would welcome any comments explaining why the chroma shift is always to the left and bottom of each image.

David

Jeff Donald
October 22nd, 2003, 11:33 AM
Andre De Clercq' s post above just about covers it. It would be impossible to be much more specific without test instruments to analyze the video signal. Send the camera to the manufacture before the warranty expires. Good luck.

David Deehan
October 22nd, 2003, 12:14 PM
Jeff,

Thanks for the very prompt reply. I read Andre de Clercq’s reply some time ago but sadly don’t have such exotic equipment. Having read Andre’s posts in other threads makes me realise how limited my technical knowledge is.

Anyway, could you advise me if the symptoms I am experiencing are normal with XM2’s (or GL2’s), is a possible defect or operator misuse? I hate the thought of packing, insuring, posting and waiting only to be told by Canon that the camera is within factory specification.

David

Jeff Donald
October 22nd, 2003, 12:18 PM
Some of the shots may have a bit excessive color bleed. But it is really impossible to tell without the correct test instruments. If it were my camera I'd send it in. Be sure and enclose a tape with a couple of clips showing the color bleed issue.

Chris Fangio
October 22nd, 2003, 01:09 PM
I don't wanna be offensive, but I doubt that any of the XM2 users (and most probably the GL2 users, too) do NOT have this sympton of the bleeding colour.

I own a XM2, I worked with a friend's XM2 and we have two units in our advertising agency. Big surprise: all of them do have the colour-bleed-problem. We tried different setups but the bleeding of red colour is clearly visible. I'm almost sure that it's by design. This and the fact that you ALWAYS have to crop the image when doing picture-in-picture video or web-video with a PAL XM2 (remember the flickering line on top, we discussed the problem here) is very disappointing for me. Please don't mention the "wanted degradation" of the still image capabilities or I'll start crying. Don't get me wrong, I like this cam (beacuse I paid 2100€ for it) but I think that Canon could have done some things much better with no expense.

You'll have to live with it or buy another camera.

Chris F.

David Deehan
October 22nd, 2003, 01:50 PM
Hi Chris,

Could you confirm that the symptoms you experience are the same as mine, i.e. nearly 100% of all colour leakage is to the left and bottom of coloured areas. I notice very little to the right and top. Do you also experience colour bleed in blues and greens?

What annoys me and what convinced me to buy an XM2 (apart from reviews on this site) was a review in the “Camcorder User” magazine - UK edition - dated Dec 2002. A quote from this review is as follows “there’s not a hint of colour bleed, even on the usual suspects such as bright reds and blues”. Maybe they got a handpicked camera from Canon or maybe the Advertiser/Magazine conspiracy really does exist.

Both colour leakage and the necessity to crop all clips in post (PAL version only) have convinced me to change camera when the finances allow.

David

Chris Fangio
October 23rd, 2003, 12:44 PM
David,

as images say more than words, I'll try to capture some still images. Hope I'll make it on Saturday or Sunday to post some pictures.

Chris

David Deehan
October 23rd, 2003, 03:34 PM
Chris,

thanks in advance for taking the time to prepare some images for inspection. I'll hold off contacting Canon until after.

regards,

David

Chris Fangio
October 26th, 2003, 12:00 PM
David,

I checked my captured files and found out that I have only frame-mode shots on this computer (and I saw you were using normal mode). But I remember shooting interlaced last week (fruits for a commercial). I can check this tomorrow in our agency if you want me to.

I also rember that someone posted frame grabs of his GL2 and this effect was clearly visible. I just don't remember his name (the pictures showed african people with colourful clothes if I remember right).

Chris

Paul Doss
October 26th, 2003, 02:54 PM
This is really sad! I have been saving for a long time to get a GL2. I have seem this with footage that was captured from VHS. The reds and sometimes the dark blues bleed or ghost. Filters in post help a little but, very little.

I thought until reading this thread that wouldn't be a problem with digital source material. If I'm reading this right everyone has it unless the final product goes out to MPEG2. Seems the XL1s, and some Sonys have the same problem. Makes you wonder what cameras don't have it. Is it something we just have to live with? Or is there a camera that doesn't have it.

Andre De Clercq
October 26th, 2003, 03:43 PM
David, all color encoded (PAL NTSC) footage generates some extra color bleeding because of the ectra analog color bandwidth reduction. The point is that the bleed is generally acceptable when it is not shifted w.r.t. the luminance picture. Once the bleed (resulting from DV sampling limits AND analog bandwidth limits) is shifted, the "shortcomming" comes to one side and becomes much more visible. In PAL, the extra vertical color bleed (allways downshifted) is, together with the further halved vert color resolution (two times!), one of the PAL shortcommings. NTSC only has the DV sampling limits and should apart from some color bleeding, not show vertical color shift. Like I already wrote in this thread, component playback should solve the shift problems, unless the DV codec itself would introduce some shift effects. On the XM1 which I was evaluating (and didn't buy) some time ago I reported the color shift problem (see my message on Dec 26th 2002).

David Deehan
October 27th, 2003, 09:04 AM
Replies to the last 3 posts.

Chris, I would be grateful if you could post sample images of the commercial you shot last week.
Regarding the frame grabs posted earlier, would this be the link you are referring to: http://homepage.mac.com/bhardy3/PhotoAlbum8.html

Paul, don’t be too disillusioned. The colour bleed problem only crops up now and again. It doesn’t occur in every shot. Just avoid shooting saturated reds, blues and greens as much as possible.

Andre, I think I would need a degree in television electronics to fully understand your reply. Am I right in assuming that the colour shift you are explaining can be generated in the DV encoding process (by the camera’s hardware codec) as well as during decoding (i.e. playback). In my case colour shift is evident in the playback of some unedited clips both on desktop computer and television. It is also visible on a laptop (with a tft display). Would I be correct in thinking that a laptop display is purely digital (i.e. no digital to analogue conversion) and would therefore be equivalent to using the DV component player and component display you mentioned in a previous post. If this is true would I be correct in assuming that the colour shift I experience originates from the camera.
I just looked at your post dated Dec 26th 2002. It’s a pity nobody replied. I guess most people (including myself) didn’t understand its meaning or importance. What I find annoying is that my previous digital cam did not exhibit colour shift and it cost less than half the price of the XM2. Chris Fangio seems a little upset as well (see above post).

David

Andre De Clercq
October 28th, 2003, 05:29 AM
David, your assumption is roughly correct. Indeed DV codecs (HW and SW) can show some color shift in addition to the bleeding. Your desktop playback experiment is not the best way to interprete color shift (or any DV picture artifact) because the decompression, deinterlacing and rescaling needs a "stripped down" DV image reconstruction in order to get the real time video on yr TFT screen. I think you better try to evaluate on a real component equipment. B.t.w. did you know that some DV VCR's have an Y/C delay adjustement possibility (like Sony DHR 1000..) Also some TV's and monitors have this feature (my home TV Loewe Arcada has it too). Of course this doen't solve the initial problem. Also many high end TV's have CTI (Color Transient Improvement) circuits which even solve the remaining bleeding problem for a good part.

Bernd Binder
March 19th, 2004, 03:08 AM
Hello to All,

I am using my XM2 (PAL) now for 3 months. I have also registered a colour bleeding problem (not only red) under special conditions, f. e. red dressed skiers. It is visible on TV (Sony 100Hz) and on computer monitor. The bleeding is always on the left and bottom side of the coloured areas.
I have experimented with all possibilities of the setup – I don’t work.
Last week I brought my cam to Canon Vienna for repair. After 4 days I got back the cam. The diagnosis was: Everything with this cam is within to norm. Conclusion: Learn to live with it or buy another cam.

Rob Lohman
March 19th, 2004, 06:18 AM
Can you post a full resolution frame on a site somewhere, Bernd?

Bernd Binder
March 19th, 2004, 07:53 AM
I will do it next week.

Andre De Clercq
March 19th, 2004, 05:00 PM
Bernd, the color bleed in vertical direction is inherent to Pal decoder systems wereby in addition to that color downshift, also the vertical color resolution is halved in the Pal decoder. That's one of the reasons why components is better.... If the whole video chain (camera+ display)is OK, then the horizontal bleeding should be perfectly symmetrical

Howard Meyer
March 19th, 2004, 06:02 PM
Just came across this issue of color bleeding- thank goodness. I thought mine was the only one with this problem. My GL2, purchased in December of 2003, bleeds to the right as you view the image on the screen. It also has halos around objects of different brightnesses. It has been to the factory once and returned with a note saying it is within specs. Fortunately, there was a follow-up customer survey sent by Canon. After checking 0 in all of the boxes, I received a call from Canon Customer Service. A test tape in now in the hands of their repair department and I received a phone call saying something is wrong. They have already sent me a prepaid mailer to return the GL2, again, and I am just waiting for the go ahead. Hopefully, the problem is correctable although my 30 years as a Ford engineer make me a little cautious. We have had problems with some products that just could not be fixed- and we spent millions trying.

Andre De Clercq
March 21st, 2004, 09:15 AM
Color bleeding in general is defined (limited) by the system (DV...PAL, NTSC..) and can't be reduced with camera settings or design tradeoffs.. Horizontal bleeding symmetry however is a very well known and understood issue by color encoder/decoder designers. It's been defined as the Y/C delay matching and is allmost allways selectable in the encoders as well as in the decoder(monitor or tv). In some cases it's user selectable even in consumer units. In modern TV design several kinds of color transient improvers are being used (CTI...) wereby the system limited color transient values get artificial enhancement resulting in allmost no bleeding for saturated color transients. If you want to (subjectively) verify your video chain (camcorder+display), just shoot a white paper with a red piece of tape from top to bottom and see on your screen if the bleeding is simular on both side of the tape image.

Howard Meyer
March 22nd, 2004, 03:53 PM
Andre
I photographed a red, green and blue tape against a white background and although all showed some bleeding, the red was the worst. And there was a difference from one side of the tape compared to the other. As viewed on the TV screen, the left side of the tape had a very bright fringe or halo with a reddish tint. The right side of the tape had a smeared reddish cast in its halo against the white background but lacked the brightness apparent in the left side.
Canon has had my GL2 and after returning it ask me to send in another tape showing my complaint again. Their reply was "it is within factory specs" (which is what the first reply said).
Because the GL2 has had good write-ups, I can only conclude the quality of the GL2 camcorders runs from excellant to mediocre. I hope potential GL2 purchasers buy the camera from a store that allows a return- not just an exchange. I you get a lemon as I did, the factory warrenty is vaporware.

Bernd Binder
March 24th, 2004, 03:47 AM
Here are the promised frames showing the colour bleed on my XM2.
Please follow the link http://temp.projektstudio.com/
As I have already mentioned, it’s the same on TV and computer monitor.

Howard Meyer
March 24th, 2004, 10:44 AM
Bernd
I wish the guys at Canon had eyesight as good as we do! Yes, I see your bleeding problem and my 70 year old wife with thick glasses also sees it. Very similiar to what I get on my GL2. Fortunately, Canon agreed to revisit my GL2 problem, so I am returning it to the factory for the second time. I seem to have gotten every excuse under the sun for what I see. My fear is they do not know how to fix it.
I wish other GL2 or XM2 owners who DO NOT have bleeding would respond to let us know if the problem is isolated to a few cameras. Better yet, a response from owners who HAVE bleeding problem would also be welcome. If it a universal problem we may just have to bite the bullet and live with it.
Oh, by the way, I saw a few halos in your photos in addition to the bleeding.
Howard

Steve Hancock
March 24th, 2004, 02:13 PM
I've had a GL2 for ~2 years - the first was was returned after a couple of months because of tape noise (while discussing it at my camera shop I was I was standing beside the Canaon rep - he authorized an immediate replacement). Most of the work I do is for sports videos (football) - bright colors and strong sunshine. Both cameras exhibited the bleeding/halo problems talked about here and elsewhere. I've used all the camera adjustments and filters discussed, with reasonable results. But as before, the tape playback looks good, it's OK on the computer, halos etc. visibile on a TV, and a copy to a VHS tape was terrible.

Over the last few months I've been making DVD's. Everything that looked terrble on a VHS tape is almost perfect on a DVD. So while we probably agree that the GL2 (and other digicams) have issues with bright colours (reds especially), I'm wondering if a major part of the problem is converting the digital signal to analog for playback on TV or VHS. Any comments from those electronically inclined (I'm a petroleum engineer, and didn't do well in electricity)?

David Deehan
March 24th, 2004, 03:09 PM
As the thread starter I think it’s about time I rejoined this discussion. It is probably more accurate to call the problem chroma shift than colour bleed; i.e. colours are shifted left and down. The vertical downward shift seems to be strictly confined to Pal users only (not an NTSC GL2 problem). I have had a number of emails from XM2 owners whose cameras exhibit this phenomenon.

I can concur with Steve Hancock’s finding that encoding to DVD and playback on television (using RGB input) reduces (but does not eliminate) the leftward colour shift. It does not however reduce the vertical shift. Again, this vertical shift is an XM2 problem and will not be evident on NTSC GL2’s.

Recently I got my hands on a digital projector. This was connected to my DVD player using component inputs/outputs. The shift was still evident.

Last November I sent my XM2 to Canon for inspection. The following is their reply:
“No fault found. Examined customers tape with complaint. This was found to be within specification. All camcorders will experience some bleeding, even broadcast.”
In my covering letter to Canon I did not mention bleed – all references were to colour shift.

I’m convinced this is a generic defect with all XM2’s. Therefore us poor Pal users will probably just have to suffer as its unlikely Canon will come up with any remedy so late in the product cycle.

Howard Meyer
March 24th, 2004, 03:37 PM
David
It may be a chroma shift that's confined to PAL/XM2 cameras but the GL2 has a big problem also, what ever the name. I see color spilling onto the background regardless of whether I am playing the original tape or a DVD made from the tape but my shift is to the right (when viewing the TV).
And by the way, the Canon reply must be canned- I got the same wording when I sent in my GL2. In regard to broadcast cameras having the same problem, give me a break. I have never seen anything on TV that compares to the bleeding on my GL2- not even on America's Funnest Home Videos.
Howard

David Deehan
March 24th, 2004, 04:15 PM
Howard,

I can understand your annoyance. Since the problem you are experiencing is slightly different to that suffered by XM2 users and not having seen much GL2 footage I'm not in a position to make an informed comment. Not many GL2 users have replied to the main thrust of this thread, which may mean that your situation is not symptomatic of GL2 cameras. This could indicate that your problem is fixable. Wait for a while and see if other GL2 users can confirm that their cameras do not exhibit what you are describing. If this is the case then your GL2 is probably defective - it will also strengthen any future argument with Canon.
Our XM2 problem would seem to be generic; therefore no quick fix is available.

regards

David

Cosmin Rotaru
August 5th, 2004, 08:25 AM
the same problem here - XM2 user (just wanted to add me to the list...)

Did someone tried to solve that in post? Maybe in Vegas? (as I use that!) :)

David Deehan
August 5th, 2004, 04:50 PM
Cosmin,

So you suffer the dreaded colour bleed (chroma shift) problem as well. I’m surprised there haven’t been many more XM2 users adding their names to the list. Did you see my recent thread at:

http://www.dvinfo.net/conf/showthread.php?threadid=29688

This thread describes another aspect of the chroma shift problem. I had originally thought it was some type of chromatic aberration but Andre de Clerq convinced me otherwise. The thread was started in this group but was quickly moved by Ken to the Open DV Discussion group – you may have missed it.

David

Bob Harotunian
August 6th, 2004, 08:55 AM
Read this post with some interest and decided to look for a red on white example of bleeding with the GL2. This is not scientific, but the example I posted below does not show me any significant bleed and doesn't raise a concern from my perspective at least with the GL2 model.

http://www.beaconvideo.net/temp01.htm

Andre De Clercq
August 7th, 2004, 04:22 AM
Bob, this is indeed a pretty symmetrical bleed which is fully acceptable. The problem often occurs when the bleed is no longer symmetrical due to chroma shift...all the bleed occurs on one side instead of being distributed (and thus halved over the two edges) of a colored object. Playing back in NTSC(PAL) generally adds extra bleed and sometimes Y/C shift too. Some high-end TV's have CTI (color transient improve) circuits implemented which "cut" color bleed.