Dude... if you look closely you can see the whole Oly image is shifted slightly blue
Nonsense. I just posted three samples showing significant color distortions in the Oly picture. The first one is noticeably red; the one on the edge of the box at lower left is cyan, and the one on the woman's head is purple.
probably due to minor processing differences in the original raw files.
Yeah right. Now you are REALLY grasping at straws!
In closing, you are certainly entitled to your opinions about the vast superiority of the CZ...
Thanks, that's very kind of you.
But speaking for me myself personally, nowhere in these images do I see enough of an optical superiority to validate the difference in these lens' respective market values...
IMO, the only reason you even came back here talking about the Oly lenses is your interest in their "respective market values".
I already said a couple of times that in terms of performance/price the Oly is very attractive. However, the claim that the Oly is close to the CZ irrespective of price requires one to simply ignore all the evidence that shows the contrary.
Okay, I've said all I'm going to about this and am out of here on this thread.
It would have been nice if you had apologized for calling people here "lemmings" before you left, but whatever.
flatdraft wrote:
i'm starting to worry about the both of you (jack and charles) since your argument has become a battle over what's near some old guy's butt.
-erik
I'm worried about ME - I didn't know that was a guy's butt.
10Tenths wrote:
My perception of these people is that the more expensive their equipment, the more entrenched their opinion of that equipment’s superiority.
In psychology they call it the law of commitment and consistency (Influence – Science and Practice, by Robert B. Cialdini).
It basically boils down to the fact that no one (Nikonian and Canonites incuded), wants to admit bying (in this case) inferior equipment…
mike, guy, son, and other zeiss 21 users... have you had the same results with moustache distortion?
zeiss 21 f/2.8
canon 16-35L @ 21mm
i know that in many landscape scenes this would likely not be noticed, but in a scene like in guy's cockpit photo, did this not pose an issue? is it correctable?
Yes, I have seen the "moustache distortion" in my testing of the Zeiss 21. Because of this distortion, I think the Zeiss 21 is much better suited for landscape work than it is for architectural work.
Every lens has it's pros and cons. The Zeiss 21 excels at edge-to-edge sharpness (even at wide-open apetures), micro-contrast, and relatively low CA.
The Zuiko excels at edge-to-edge sharpness (at f5.6 and higher), and very low distortion. The Zuiko generally suffers more light fall-off than the Zeiss as well, and therefore has lower contrast in the corners (which is sometimes confused as lack of sharpness, but it's actually still very sharp).
With the distortion - it is correctable, but at a price. You can use PTLens to correct (use it on a TIF or in the new release of Bibble on your RAW files), but you will lose sharpness, as the tool has to "distort" the corners to make them appear straight. I've done the correction on some images, and the corners get soft as a result of correcting the distortion.
So there is no perfect lens - they all have trade-offs. When confronted with an architectural situation like the one above, my personal favorite choice is a PC (perspective control) lens such as the Zuiko 24 Shift or the Zeiss 35 PC. They tent to have miminal distortion and hold straight lines very well.
The Canon 16-35 also has very low distortion. In fact, that's the reason that I chose to keep mine over a 17-40L (which has much higher distortion). I like the Canon 16-35 because it's a nice balance of sufficient sharpness (when stopped down) paired with low distortion. CA is an issue, depending on the subject. And the AF is nice.
Thanks again Andy - I think this is an important subject.
Andy wrote:
yuh..... what i was attempting was apples to apples, z21 vs 16-35@21mm
I understand - wasn't critiquing the comparison, but noting a characteristic that I've noticed with the lens. Yours is a valid distortion test, and shows that the 16-35 does well in the distortion area in the mid-zoom range.
charlesk wrote:
15 seconds in Photoshop using PTLens.
Not perfect, but a more controlled subject would be needed to entirely correct the distortion. --c
hi charles-
i'm aware of these fixes... but i wonder, about compromise of image quality, as mike hatam pointed out in his reply? also, if i don't have to run such a program (ptlens) i'm happier, there's enough complications in my life already
PTLens uses panorama tools, which incorporates a number of fairly sophisticated interpolation programs. They are extremely efficient and result in almost no image degradataion. In fact, they are assessed by using them to transform a whole bunch of times and then compare to the original and you would be amazed at how little loss there is.
I agree that avoiding a step is avoiding a step. Though PTLens can be batched... --c
That's a torture test that shows what happens to an image that is transformed 36 times through 5 degree rotates.
I believe PTLens uses "Poly 3". The author claims there is no noticeable affect on quality so he didn't go any higher. Other tools let you change the interpolation method.
For reference the original test image and the 36-rotate images follow.
I spent some time writing code to interpolate Canon CRW files, as decoded by Dave Coffin's utilities. The writeup on lossy rotation by Helmut Dersch is misleading, and lead me on a wild goose chase. The Fourier Transform is lossless. The Sinc function (usable for image interpolation) is lossless, but yields an ugly result (ringing on high-contrast edges). Sinc function with Lanczos windowing is better, and lossy. All around-best is probably cubic spline. In image processing, one should prefer a pretty result, not an invertible function.
In choosing a lens for the digital domain, certain image defects are essentially uncorrectable (longitudinal CA, coma), while others (transverse CA, linear distortion) are correctable without serious concern for image degradation. Highest quality will come from a single-pass application of all desired image transformations: Bayer interpolation, correction of linear distortion, rectilinear reprojection after cropping, correction of TCA, and image resizing.
Based on your claims, I think that the new Bibble (with PTLens incorporated into the RAW processing) might be the best way to process distorted images. Unfortunately, Bibble has been a bit behind RSE and C1 in generating fine detail during RAW processing. But maybe for distorted images, it's the best route for processing.
Anohter thing that isn't obivous to most folks is that doing such things in linear space is the prefered mehtod. PanoTools, the underlying library for a lot of things acutally attempts to map the image data from gamma corrected space back to linear space before performing many of its corrections.
Once an image has gone thorugh a conversion pipeline, error has been distributed further through the image, and if you are dealing with 8 bit images, detail has already been lost in the ihglights through gamma and tone compression.
The new BPTLens plugin for bibble works at th earliest point in the image pipeline on linear data to help do things at the highest quality. This should yield even better results once we hook up the BPTlens CA filter in a future version.
Can any of you who obviously know far more than I do about this subject quantify in any way the loss in detail that results from a single interpolation/transformation? Is this possible? Or are there examples showing the deleterious effects of same?
As someone who uses PT a lot but has not noticed any degradation that I could detect, I would be most interested in learning more. Thanks. --c
Eric - thanks for making an appearance in this thread, and for your innovative work to bring distortion and CA correction further upstream in the workflow process (into the RAW conversion itself).
