Chromatic abberations

[unpopular]

In particular zoom lenses often show a considerable amount of lateral CA, even the good ones. Luckily lateral CA is quite easily reduced if you start from the RAW image, without any loss in image quality. Actually the image in the different colour channels are shifted laterally into the right place, and this increases overall sharpness!

I am unsure why raw processing would be less invasive than even 8-bit image processing for lateral CA, if all that is happening is the channel information is being shifted. Simple translation never affects image quality.

It is entirely likely that CA is being corrected after the raw image is processed anyway.

--

Because CA is by nature monochromatic, one can use local hue/sat correction with a very narrow feather to minimize the noticeability of stubborn CA.

[Ads]

[sovietdoc]

I used to have a ti1 with 17-85mm and CA was clearly visible with that lens. Here on 5d3 and 70-200 ii I've TRIED to make shots so thay would have profound CA, I've never ever seen it.

[unpopular]

That's an interesting observation, and kind of makes sense. Relative to the sensor, the CA is smaller. If I'm thinking clearly, it's there, but occupies fewer pixels.

There is a chance also, though I kind of doubt it, that the AA filter may also introduce some CA, if the AA filter on the 5D3 is thinner and more precise, then less CA would be present. But again, that's just a baseless hypothesis.

[Garbz]

I don't think so. CA is present when light is bent through an optic medium. I don't think you can get it when propagating through a rectangle as the entry and exit angles are complimentary, so a AA filter would not contribute to CA. If it does it certainly wouldn't be field relevant.

[Helen B]

I don't think you can get it when propagating through a rectangle as the entry and exit angles are complimentary...

Although the entry and exit rays are parallel, rays of different wavelengths can emerge from different exit points even though they had the same entry point. This causes axial CA, albeit very slight (usually).

[Dao]

I believe the refractive index varies with wavelength, so I think the AA filter should have effect on the different wavelength of light (different wavelengths have different exiting angles). But is it enough to cause visible difference, then I am not sure.

[slackercruster]

Do chromatic abberations matter? Can you fix it in post without reduction in image quality?

If you got a fine shot try to fixe em. Sometimes CA are nice to showcase though...

[unpopular]

I believe the refractive index varies with wavelength, so I think the AA filter should have effect on the different wavelength of light (different wavelengths have different exiting angles). But is it enough to cause visible difference, then I am not sure.

Can birefringence take place at different wavelengths? I am guessing so, and that this is why you see rainbow patterns. I kind of have a hard time seeing an AA filter splitting light into perfectly neat packets of monochromatic light. Might this be why you sometimes see color fringing on fine details?

While I don't think that an AA filter alone would create significant amount of CA due to it's thin size and immediate proximity to the sensor, if it either refracts one spectrum or the angle of birefringence is different at different wavelength, it may make CA more pronounced. I think that this is a pretty long shot, and I doubt that any increase in CA due to an AA filter would be noticeable.

[Helen B]

I believe the refractive index varies with wavelength, so I think the AA filter should have effect on the different wavelength of light (different wavelengths have different exiting angles).

I'll rephrase what I wrote earlier. Wavelength does not affect the exit angle (from a parallel-sided block or filter) but it does affect exit location. It is the difference in exit location that causes CA, though it is unlikely to be perceptible in most cases.

[unpopular]

So then it wouldn't matter the distance from the sensor then? Basically the CA affected wavelengths are simply transposed relative to their entrance upon exit?

[TCampbell]

CA is fairly easy to correct in software. The camera sensor records the image onto the sensor using a bayer mask. So the red, green, and blue are all recorded on their own "photo sites" and later combined to create "pixels".

Light refracts through the lens and, just like a rainbow, the wavelengths of light start to split into their constituent colors -- just a tiny bit (achromatic doublets in the lens try to minimize the effect but cannot cancel it entirely.)

Imagine you separate an image to create a "red channel", "green channel", and "blue channel" version of an image. Then imagine shrinking the blue channel image by about 1%, growing the red channel image by about 1% but leaving the green channel image alone -- then recombine back into a single image. I'm making the percentage up -- but that's the general idea.

Since CA varies by how far a point is from the center of the image circle projected by the lens, the very center should appear to have virtually no CA at all... while the corners should have the most CA. This means that if you perform the correction right out of the camera, you can then do other things to the image with impunity. But if you start to resize and crop in sections of the image first (say you've cropped into a detail in the lower right corner) and THEN try to apply a CA adjustment, it may not work because the center of the image you see (post crop) was NOT the center of the image circle being projected by the camera lens when the shot was created.

Part of the point of the camera having a built-in UV & IR filter is because those wavelengths, being at the extreme ends of the visible spectrum (but still visible to the sensor -- just not to human eyes) would cause even _more_ CA. By cutting IR & UV light, the image is sharper.

[Helen B]

So then it wouldn't matter the distance from the sensor then? Basically the CA affected wavelengths are simply transposed relative to their entrance upon exit?

Yes, in terms of any (very slight) longitudinal (axial) CA caused by the AA, it shouldn't matter.

CA is fairly easy to correct in software. The camera sensor records the image onto the sensor using a bayer mask. So the red, green, and blue are all recorded on their own "photo sites" and later combined to create "pixels".

Light refracts through the lens and, just like a rainbow, the wavelengths of light start to split into their constituent colors -- just a tiny bit (achromatic doublets in the lens try to minimize the effect but cannot cancel it entirely.)

Imagine you separate an image to create a "red channel", "green channel", and "blue channel" version of an image. Then imagine shrinking the blue channel image by about 1%, growing the red channel image by about 1% but leaving the green channel image alone -- then recombine back into a single image. I'm making the percentage up -- but that's the general idea.

Since CA varies by how far a point is from the center of the image circle projected by the lens, the very center should appear to have virtually no CA at all... while the corners should have the most CA. This means that if you perform the correction right out of the camera, you can then do other things to the image with impunity. But if you start to resize and crop in sections of the image first (say you've cropped into a detail in the lower right corner) and THEN try to apply a CA adjustment, it may not work because the center of the image you see (post crop) was NOT the center of the image circle being projected by the camera lens when the shot was created[*].

Part of the point of the camera having a built-in UV & IR filter is because those wavelengths, being at the extreme ends of the visible spectrum (but still visible to the sensor -- just not to human eyes) would cause even _more_ CA. By cutting IR & UV light, the image is sharper.

That only applies to lateral CA, not to longitudinal (axial) CA which shifts the focus plane. That is more difficult to correct.

*The issue with images that do not have the lens' optical axis passing through the image centre happens all the time with lenses that have been tilted or shifted. You have to enlarge the image with blank space to place the lens axis at the approximate centre of the new image size before doing software CA correction.

[unpopular]

Helen, I was reading about calcite lenses in trilobites (lol, yeah) and wandered over to this article:

Gradient-index optics - Wikipedia, the free encyclopedia

Might this technology find it's way into imaging lenses to help minimize CA? Could the lenses of the future be a solid piece of glass with variable index of refraction throughout?