Distortion correction 14-35 RF lens

Started 4 months ago | Discussions thread
the-genuine-particle New Member • Posts: 24
Re: Distortion correction 14-35 RF lens

Karl_Guttag wrote:

Tim 6789 wrote:

Am I understanding this right? At 14mm, the RF 14-35 distorts heavily unless corrected in camera or in software. It appears that the corrected image is cropped maybe by around 10% to create a jpeg or a corrected raw (source the Digital Picture lens review distortion image at 14mm). On a R5 the pixel dimensions are 8192 x 5464, so a cropped corrected version maybe only approximately 7370 x 4915. Are the actual pixel dimensions of the final image at the advertised size? Has the cropped image been upscaled to produce the final image at the standard pixel dimensions? Does this also apply to the 24-240, which also has a heavily distorted uncorrected image at the wide end? Does it apply to any other RF lenses? If so, does it matter?

First, no, and perhaps ironically, DPP4 with highly distorted images, they output MORE pixels than are on the sensor. A key thing to understand is the difference between a "photosite" (photodiode with a single color filter over it) and a "pixel" with all 3 colors.

Just to get this out of the way, ALL digital camera images are heavily processed. It starts with the RGBG Bayer filter on the sensor to pixel conversion. Full-color pixels are counted as if each photosite has all 3 colors when in fact, they have only one. A high-quality RAW conversion looks not just process 3 to 4 nearest photosites but many neighboring sites (I would guess on the order of a 5x5 photosites. In short, you don't want to see the sausage being made.

So dealing with distortion just has to factor this into the mix. What you don't want to do is first convert RAW to RGB and then do distortion correction, in which case you will get more substantial resolution losses.

There is not a free lunch. The outer parts of the image with significant barrel distortion correction will lose some resolution. Effectively, more of the image is concentrated in fewer photosites. But the center of the image will lose very little, if anything.

Fundamentally, RF lenses seem to be designed with digital correction in mind. But then so are almost all modern lenses. Lens designers can make trade-offs between what they fix in the lens design. Geometric distortion is relatively easy to correct. So are some types of chroma (color) aberrations. What you want is that the light of the various colors is focused well. Some level of vignetting is also easily correctable with some increase in noise.

I also have Olympus equipment, and they don't even have a mechanism to see an image without distortion correction.

Overall, Canon RF lenses, even the wide-angle zooms, look very good after correction. The highly distored ones lose some resolution in the corners, but the contrast and chroma aberration are very good after correction.

In the "modern era" you really should be comparing the "post corrected" results. A lens that may look better pre-correction, might not be as "correctable." I see this when I have compared EF to RF lenses.

The biggest downside of all this manipulation is that you lose the natural look of the optical falloff. The corner softness on the RF 24-240mm tends to have a disturbing looking smeary appearance compared to say the Sigma 12-24mm which looks natural in the corners.

Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow