CornerFix - curious results

[Guest V64]

Advertisement (gone after registration)

I have been trying to code a CV21/4 (as variously TriElmar, ASPH, PreASPH) but all the images I get show a faint but distinct Magenta cast across the bottom edge and corners (as though they are over-corrected).

 

So I tried CornerFix - with much the same results.

 

My first reaction was that my Profile image was 'not good enough', so I used an illuminated White Box as a source, which made little difference.

 

So as an experiment I Cornerfixed the Profile Image using the profile generated from that image, and then used the Auto Levels tool in C1 v5 to 'exaggerate any color casts or deviations in the image. The results are 1. The CV21/4 2. Cron28mmASPH, Elmar50mm2.8. All with filters an no lens detection. Please ignore the reflections of my fingers/camera - these just show how sensitive the process is, and how difficult it is to get a good profile image.

 

What do these images say about the lens and/or Cornerfix?

 

I look forward to your comments

[sandymc]

Difficult to tell from those images - can you post the original DNGs somewhere?

 

Sandy

[jaapv]

Whatever else, you need to clean your sensor

[Guest V64]

Sandy,

 

I have just found Cornerfix V3 (took a bit of finding!). So, first, I think I should apologize for my first post, it is certainly out of date.

 

The 3rd version looks to have dealt with the 'decentering' which I think was the issue for the Cv21 which possibly also effects the in-camera corrections when used coded.

 

The V3 images for the CV21 are much more uniform - even under the extreme processing of C1 v5 auto levels.

 

However the CV21 images remained rust coloured' and the Leica images remained 'green'.

 

So I white balanced the processed images before doing the auto levels the differences between the CV21 and the cron 28 essentially disappeared into a kaleidoscope of noise and diagonal patterns.

 

So maybe CornerFix V1 a good way to test for lens decentering!

 

I will explore some more, e.g using better control (Manual) White Balance of the profile image, and hopefully I can post some revised findings.

[Guest V64]

Whatever else, you need to clean your sensor

 

It would seem so - but none of the spots appear in any 'real life, correctly processed images' - the C1v5 auto levels is a very aggressive tool

[Guest zoz]

I don´t use C1 but usually auto-level-tools are rising contrast to a level that the histogramm is filled. So the original image might be as even as possible but the auto-level will bring up some structure.

Imho it is absolutely impossible to judge about evenness of the file after using auto level.

[Guest V64]

Advertisement (gone after registration)

Update

 

I have now reshot the Profile images for the CV21 and a cron28 using the whitebox and Manual White Balance, and processed using CornerFix v3.

 

This approach removed the differences in color, which, it would appear, may have been due to different Auto WB (CV21 - 4988, Cron 28 - 4594, Elmar 50 - 4554, Manual WB 4683) between the various lenses.

 

So the messages for me are:

1. use Manual WB on any shots which have to be processed by CornerFix, especially when taking Profile images.

2. misaligned lenses show slightly uneven corrections when done in camera (maybe no surprise).

3. CornerFix V1 (process the profile image with its own profile and view in C1 v5 with auto levels) may be one cheap way to check a lens for alignment.

 

Thank-you all for your comments, and yes I have cleaned my sensor (another use for Cornerfix and the lens check process).

[sandymc]

Ok, you should be using V1.3.0.8

 

The most recent version or CornerFix can always be found here: CornerFix | Download CornerFix software for free at SourceForge.net

 

Sandy