Canon HF-G10: Chromatic aberration

[Bryan Worsley]

I recently acquired (EBay auction) a Canon Vixia HF-G10 - my first AVCHD camcorder, in fact. One of the reasons I chose this camcorder (budgetary constraints aside) was in the light of comments made in several product reviews (supposedly independent) that the HF-G10 displays relatively little CA (“fringing”, as it is often mis-termed, apparently), at the telephoto end, by virtue of the (then) new aspherical lens.

I’ve had the HF-G10 a few weeks now and while I have been very impressed with the resolution and dynamic range (low light capability included), I am a little concerned about the degree of CA I have been seeing at full wide angle.

With that, I took it upon myself to do a little testing with a CA Distortion Test Chart. Nothing fancy, just printed one off the web on A4 paper, stuck it on the wall and recorded short clips (on a tripod) at various points over the optical zoom range, repositioning the camcorder at each point to fill the screen. But it served it’s purpose. I’ve put up a short video on Vimeo showing some of the results:

Canon HF-G10 Chromatic Aberration Tests on Vimeo

As will be seen from the chart images (at least in the second series of magnified images), CA , manifesting as classic bi-lateral purple/yellow chroma shift, was present to some degree over the entire optical zoom range. The purple form generally appeared with greater intensity on the left side of the field, whereas the paler yellow form dominated the right side. As expected ,CA was strongest at the edges of the image (Note - the green strip on either side of the chart in some of the images was the wall)

At maximum zoom (x10 optical) the CA was relatively diffuse with the purple and yellow shifts emanating from opposing edges of the black stripes - fairly typical. What surprised me was to see an (arguably) comparable degree of CA at full wide angle, the difference being that the shifts were more tightly associated with the black edges; presumably this relates to the sharper resolution of high contrast at wide angle. Furthermore, the CA did not immediately disappear on zooming out from wide angle. Rather it decreased progressively down it’s weakest level at around 50% zoom, stayed at that level up to about 80% zoom and then increased in intensity from 90% to maximum zoom.

The CA chart images pretty much reflect what I have been seeing in wide angle shots. Here are two typical examples - frame grabs from clips shot indoors at 1080/30PF (Program AE mode, +12db max gain, AWB). Not exactly aesthetic in terms of subject matter - but hey, that’s family life.

Note: to view the linked images at original resolution - click the 'enlarge' button (bottom right) on the first opened image and then again on the enlarged image.







In the first image (with my son testing a new ‘diabolic’ image stabilization system) one can see (hopefully) the ‘purple’ type chroma shift, more on the left side of the frame - for example, on the edges of the white paintwork and the blue stair carpet, and on the left side of the arm on the beige arm chair. And then the 'yellow' type, appearing more on the right - on the back of the dining room chair, for example, where the padding meets the wood frame.

In the second image (with my son assembling a DIY in-camera CA compensation kit) - more of the same. Granted, this was shot with an household halogen light shining on from the left side, so it was mixed with window-daylight - but, hopefully, one can see, for example the bi-lateral CA on the edges of his red shirt - purple on the left edge, yellow on the right - as well, as more of the 'yellow' type on the white paintwork to the right.

Now I appreciate that some degree of CA can be expected in “sub-professional level” camcorders. What I’m struggling to gauge is whether this degree of CA, at wide angle, would be considered normal for the HF-G10 or whether it is something I should be concerned about.

Would appreciate any feed-back, especially from other HF-G10/XA-10 owners.

[Bryan Worsley]

Strange....can't seem to open the Vimeo video from the post. The URL is correct - double checked. What could be wrong?

[Nate Haustein]

This looks normal. I just tried it out on my HF-G20 and it was pretty much the same deal. Purple fringe on left side most prominent at full wide-angle. I wouldn't worry - you don't have a bad copy and your home videos look great.

[Bryan Worsley]

I'm still not sure why that Vimeo video is not showing, but, anyway, thanks, that's good to know. Like I said, this is my first AVCHD camcorder.

I did in fact test out an HF-M400 (the HF-G10's poor cousin) a while back and one of the reasons I returned it (aside from the missing stock battery) was the very obvious 'yellowing' I was seeing on the edges and contours of pale objects - paintwork, around doors etc. Looking now at the test clips I kept, I realize that it was the same type of yellow chroma shift only more severe. Didn't look too bad on a PC, but displayed on an HDTV it took on a horrible 'ghoulish green' type of appearance. So, what CA I'm seeing on the HF-G10 is definitely much, much less than that....and so it should be.

Strangely enough, when comparing with my Canon HV30, it would easy to suggest, at face value, that the HF-G10 shows more CA. After running a parallel series of CA chart tests with the HV30, however, it is clear that CA is there also, its just more diffuse, even at widest angle - so you see less of that 'tight' buzzy chroma fringing on high contrast edges and more of the diffuse-type on lower contrast gradients...lower resolution of course, as well.

Anyhow, CA aside, I think the HF-G10 is a great camcorder, and it well suits my needs.

Cheers.

[Bryan Worsley]

By way of follow-up, I wondered if closing down the aperture might reduce the degree of CA seen at maximum WA, as 'conventional' lens theory might suggest.

So I repeated the CA Distortion Chart tests again at 30PF, under pretty much the same lighting conditions as the first tests - ambient (window-lit) indoor light, registering 5500K on manual balance. Set at max WA, I then incrementally closed down the aperture with the camcorder first in Aperture Priority AE mode and then in Manual exposure mode at a fixed shutter speed. Gain was again capped at +12 db....my standard/general purpose setting.

I also tested to see if a Circular Polarizing (CPL) filter had any effect, again at WA in Program AE mode.

Short answer - no, reducing the aperture did not appear to have any impact on the degree of CA seen at full WA, under these conditions, at least over the F stop range tested (F1.8 - F5.2 ND 1/8 in Aperture Priority Mode), and neither did a CPL filter.

It would be reasonable to ponder then whether the CA that is seen is occurring primarily at the sensor micro-lens level. Interesting to note that in developing the HD CMOS Pro sensor, Canon not only increased the pixel size (technically 'depth') but increased the curvature of the sensor micro-lenses to increase their light collecting capability.

Canon : Canon Technology | The Minds Behind the Magic

Whether this presented any greater or lesser challenges for CA control at the sensor level, I guess only Canon would know.

What these tests emphasized to me however is just how light sensitive this camcorder is. When performing the original set of CA chart tests in Program AE mode, I had assumed that would likely have been operating at fully open aperture, as would be the case on my old HV30 under these conditions - and any dimmer than that, I'd be reaching for my on-camera video light to keep it out of gain.

Not so on the HFG10. Although (unless I'm missing something) it is not possible to know the shutter speed set by the camcorder in Aperture Priority AE mode (as one can on the HV30 by half-pressing the Photo button), I was able to reduce the aperture down to F5.2 ND 1/8 under these ambient light conditions before the exposure warning symbol started blinking. And then in Manual Exposure mode I was only able to bring the exposure down to an usable level (as judged visually and with the Zebra markers) at fully open aperture (F1.8) by increasing the shutter speed to 1/500 sec (at +12db max gain) and around 1/100 -1/250 sec at zero gain. And conversely, with the shutter speed set at 1/60, aperture could be reduced to F4.0 (ND 1/2), at +12db, and F3.2, at zero gain and still achieve adequate image exposure under these conditions.

Not that I'm complaining about that...at all.

[Bryan Worsley]

Originally Posted by Bryan Worsley

Although (unless I'm missing something) it is not possible to know the shutter speed set by the camcorder in Aperture Priority AE mode (as one can on the HV30 by half-pressing the Photo button)

I obviously was missing something - the Aperture and Shutter Speed values are briefly displayed when pressing the on-screen Photo symbol......at the expense of taking a photo, of course.

[Steven Ansell]

Most of the purple glow you see on small sensor camcorders isn't CA, it is caused by light scattering from the sensor microlenses. You generally won't see it on larger sensor cameras or cameras that sample from a higher resolution array.

[Bryan Worsley]

Thanks, yes, I've been trying to educate myself more on this stuff.

From what I've read, the challenge for control of bi-lateral (blue-yellow, purple-yellow) CA at the CMOS sensor level derives from the fact these camcorders are tuned for correctly focusing green light, from which the luma (in YUV color format) information is derived. The microlenses serve to direct more light into the photo sensor wells, but have the potential to introduce CA (through scatter) in the process. The shape and composition of the microlenses therefore have a critical bearing on both their ‘light collecting’ efficiency and susceptibility to chroma aberrations. Probably an over-simplification, but that’s how I understand it.

According to that Canon article I referenced about the HD CMOS Pro sensor technology, changes were made to the curvature of the micro-lenses to increase their light-collecting capability, and I read (somewhere else) that they were made thinner also, presumably to increase light transmittance. So, it is interesting to speculate what challenges these modifications presented for CA control.....potential for more scatter?

What I’m left wondering though is why there was such variation in the degree of CA seen in these chart images (and footage also) at points over the zoom range. Zooming out from full wide angle, the degree of CA had only reduced a bit at 25% zoom. But at 50% zoom it had diminished markedly and at around 75% zoom it was barely detectable.

Can’t say enough about these multi-component lens and their optical paths to know how the light image is finally presented to the sensor, in terms of relative angles, but at face value, it would appear that the optics do play a major part in dictating the degree of CA.

Clearly, from a practical view-point, the notion of operating the camcorder in the 50 - 75% zoom range, purely to minimize CA, negates one of the prime benefits of this model over it’s predecessors in terms of wide angle capability.

On inquiring around other forums, I've had responses from HF-G30 users - some saying that they are seeing very little if any CA by comparison, others saying that they are seeing a similar thing on theirs. I guess it's difficult to assess without comparing both models directly.

[Steven Ansell]

I have a G30. There is slight lens CA when using telephoto. There is however the purple glow when the light is right, just like pretty much every other small sensor camcorder out there. The purple glow is not caused by the lens, you see it in professionally shot nature TV shows as well when something up in the trees is shot against a bright sky (monkeys for example). Presumably they are not using cheap lenses or cameras. The same glow is never seen in SLR photos, even when cheap lenses that clearly have significant real CA are used.

It is the sensor that does it, and short of moving to a large sensor camera you just need to shoot around it by making sure conditions are right.

[Bryan Worsley]

OK, now I'm confused; possibly a little knowledge is a dangerous thing. I was under the impression that the bi-lateral (purple-yellow) chroma shift that was evident in the CA chart test video and the two frame grabs I posted was Lateral (or Transverse) CA.

Taking this reference for example:

Chromatic aberration - PanoTools.org Wiki

Quote:

"If all of the following points are met your image most likely contains true TCA (Transverse Chromatic Aberration):

Corners should show most color fringes whereas the center should show none.
Color fringes should be not only at the edges of overexposed areas but at lower contrast edges, too.
Color fringes should be of complementary color (red-cyan, blue-yellow or purple-green) on opposite sides of a dark or bright area.
Color fringes should be in all corners the same direction pointing out from the center."

And that's exactly what was seen. Since the Chart Test video doesn't appear to be playable from within the post, here are a few frame grabs from the original test clips:

At full wide angle - note the preponderance of purple chroma shift on the left side of the image and the paler yellow shift on the right. On both sides, the intensity of the shifts increase towards the edges.



Top left quadrant of the same shot. Magnified, it is clear that both purple and yellow shifts are present to some degree on opposite sides of the black squares, it's just that on the left half of the field, the purple form is more intense.



And then here are the chart test images at 75% zoom - just the tiniest trace of chroma shift visible in the magnified image.

Full Chart:



Top-Left magnified:



I'm just wondering then if the 'purple glow' that you describe is what is often referred to as "purple fringing" or "bloom", that is - the purplish halo that tends to occur around over-exposed/blown-out subjects - bright lights, backlit windows and, as you describe, on dark edges against against a bright sky. Again, from what I've read, this is a different phenomenon to CA. With CCD sensors it is said to result from over-spill from saturated photosites. CMOS sensors are, in theory, immune to that, but a similar aberration may be due to stray/scattered UV and/or IR light.

Taking this video for example, which I know, from the producer, was shot on a Canon HF-G30; one see that type of "purple fringing" around the overhead lights. There are also light purple casts in some shots (possibly due to the mixed fluorescent lighting, WB etc), but I'm hard pressed to see much of what I'd call CA on the trims of the locomotives and close-ups of the mechanical parts.

Locomotion - Video Dailymotion

Anyhow, here are two other frame grabs from 30PF clips shot at full wide angle on my HF-G10 , which illustrate what I am referring to.

First one: note the 'purple fringing' primarily on the left side of the image - edge of the white wall, the ball next to it, left edge of left red goal post and the boy shooting with the hockey stick.



And then, from the same clip with the boy now on the right side of the field. Now, the more obvious purple fringing is replaced with the fainter yellow fringe, visible here on the back edge of the boys t-shirt.



If that's not bi-lateral (transverse) CA then what would you call it? I just can't see how one could shoot around it when it presents in typical everyday shots like these.

Anyhow, causes/definitions aside, my principal reason for posting on this subject was to try and gauge from other HF-G10 users, ideally, whether the degree of "fringing" I'm seeing is normal for this camcorder. If that's how it is, I guess I'll have to live with it. It just surprises, and disappoints me a bit, to find that, despite the advancements in image processing, my old HV30 shows comparatively little (or less blatant) aberration of this kind, not too mention fewer compression artifacts....seems that the H264 coding quality hasn't really improved much since the early AVCHD models. Still I guess it's worth it for the dynamic range/low-light capability and better stabilization.

Cheers.

[Steven Ansell]

It is just light scattering mostly. There is a little bit of CA, but you will notice that most of the color offset is associated with the halos generated during the debayering process at contrast boundaries. Probably those areas of the image are particularly vulnerable to the effects of light scattering, so you notice it more there.

Purple fringing is the most common artifact, but it can be other colors as well. Notice how the edges of the net frame have a red glow?

[Bryan Worsley]

Originally Posted by Steven Ansell

It is just light scattering mostly. There is a little bit of CA, but you will notice that most of the color offset is associated with the halos generated during the debayering process at contrast boundaries. Probably those areas of the image are particularly vulnerable to the effects of light scattering, so you notice it more there.

Ah, OK, so that's what causes the halos. I wondered if they were perhaps a by-product of the internal sharpening process. My old Panasonic GS400 DV camcorder used to be riddled with them. I tried reducing the sharpness on the HF-G10 (through Image Effects) btw - if anything, it made the fringes a little more diffuse.

Originally Posted by Steven Ansell

Purple fringing is the most common artifact, but it can be other colors as well. Notice how the edges of the net frame have a red glow?

Yes, I looked at that again, after I'd remarked on it, and realized that it was probably not the same thing. The red glow is actually more intense against the darker background shadows.

Ah well, you live and learn. Thanks for your comments.