What are some lens flaws that can be corrected perfectly (or close to) during RAW elaboration, and hence don’t matter as much as sharpness?
Or, which lens characteristics are less damaging to fine details and overall image quality, thus can be recovered when elaborating the RAW files?
As far as I understand a tiny bit of sharpness can be recovered by using deconvolution (sharpness slider set to maximum in ACR or RawTherapee), however needless to say, if it’s not there to begin with, it cannot be created without making it up.
I am asking this in order to better understand which compromises I could accept when buying new lenses, and if I’d do with a more distorted 14mm or a less distorted but more chromatic aberrated 20mm (just examples).
What about distortion, vignetting, chromatic aberration?
How would you order them in terms of importance in regards to preserving sharpness and details after post-processing?
Is distortion correction as non-deleterious as I think, or is it worse than automatic chromatic aberration removal?
Thanks in advance
1 Like
Think about what the required corrections in post-processing imply (I’ll only look at traditional Bayer matrix here).
For vignetting, we don’t have to move pixels around, only change the values => shouldn’t lose details.
Chromatic aberation: if it can be done before demosaicing, we only have to move the red and blue pixels around, so detail loss should be limited (most detail is in the green channel). If it needs to be done after demosaicing, it will probably lose a bit more detail.
Distortion removal requires moving pixels in all channels, which means we have to interpolate pixel values. That is most likely to lose detail.
For all three of these, the effects will be negligable in the center and strongest in the corners.
That said, how are you going to use your images? Downsampling will also lose you details, and probably more than the lens corrections. That’s why the camera and phone makers can get away with rather severe corrections in software, which allows them to use cheaper lens constructions: most images are only viewed on (often tiny) screens…
To get an idea of the amounts of downsampling:
I get raws of 6000×4000, for a screen of 1920×1080 pixels, so for a full-screen image a screen pixel corresponds to ~10 raw pixels). For printing, that same 6000×4000 raw would give me a full-resolution print of about 60×40 cm, which is usually watched from a certain distance… a standard 15×10 cm print is downsampled to about 1500×1000 pixels, so 1 printed pixel corresponds to 16 raw pixels. Granted, if you crop your images, you get less downsampling…
2 Likes