dxo for lens correction only before RAW processin?

[JDClements]

Thanks for the tests. The real reason I got DxO was because I am now shooting full frame, mainly with the Canon 24-105, so I figured DxO would be useful for geometry correction. In those cases, I can put up with the triple file size, but I'll be sure to turn everything else off (with the possible exception of their sharpening aka lens softness correction).

[Panopeeper]

I figured DxO would be useful for geometry correction

This is NOT against DxO or for any other raw processor (I don't have any experience with DxO). However, this needs to be said and understood clearly:

the correction of geometric distortion is not the task of the raw conversion, except if the raw conversion software happens to offer the best lens distortion.

It can be incorporated in that process, but it is nonsensical to make a separate pass of raw processing only for the lens correction, as long as any program on the path down offers an adequate lens geometric distortion correction.

It is a different issue, which software makes the best lens correction, where the geometric distortion is one of the aspects, the other being vignetting (which too is not the subject of raw conversion).

EDIT: the vignetting can be reasonably incorporated in the raw conversion; in fact, it could be done even before the raw conversion, but it can be done later as well.

[JDClements]

the correction of geometric distortion is not the task of the raw conversion, except if the raw conversion software happens to offer the best lens distortion.

That is a good point. For automated correction (with supported body/lens combinations) it may well be. However, I am also going to test it at the *end* of the workflow and see what it comes up with.

[sandymc]

You really have to distinguish between vignetting corrections and lens distortion corrections. You can happily make vignetting corrections to raw files - its a little complicated because of black levels, etc, but that's what CornerFix for Leica M8's does. Lens distortion is a different problem. The issue is that in order to perform lens distortion corrections, you have to interpolate pixel positions, which is the same as what demosaicing does. So if you separate raw conversion (demosaicing, etc) and the lens correction, you're running two different interpolation algorithms on the same image. Can be done, but not ideal. That's why DxO demosaics and corrects in one pass, and generates a linear ("non-raw") DNG.

Sandy

[madmanchan]

It can be incorporated in that process, but it is nonsensical to make a separate pass of raw processing only for the lens correction, as long as any program on the path down offers an adequate lens geometric distortion correction.

It is a different issue, which software makes the best lens correction, where the geometric distortion is one of the aspects, the other being vignetting (which too is not the subject of raw conversion).

EDIT: the vignetting can be reasonably incorporated in the raw conversion; in fact, it could be done even before the raw conversion, but it can be done later as well.

It depends ...

If you're doing other corrections as well which require resampling the image (e.g., lateral chromatic aberration correction), then it is definitely desirable to combine them all into a single rendering pass when possible, instead of doing multiple resampling stages. And some of these corrections work best in the camera's native coordinate system (strongest color separation), so it is useful for the raw converter to provide the correction facility directly. Furthermore, for some things like distortion correction, if working with the raw data, you sometimes have the ability to grab usable pixels from beyond the maker crop rectangle. Having these extra edge pixels is also useful when resampling.

[Panopeeper]

If you're doing other corrections as well which require resampling the image (e.g., lateral chromatic aberration correction), then it is definitely desirable to combine them all into a single rendering pass when possible, instead of doing multiple resampling stages

This thread started out with

I want to use the software just to make lens corrections, CA, distortion, on my RAW files, but still use capture 1 or similar for the actual RAW processing

[madmanchan]

True, but I was simply addressing the point that you made above and highlighted in bold, i.e., "the correction of geometric distortion is not the task of the raw conversion" -- by saying that there is in fact a solid technical argument in favor of doing geometric distortion correction during the raw conversion.