Focussing Monochrom with filters

[Lindolfi]

Advertisement (gone after registration)

It isn’t quite as simple. There are twice as many green sensitive pixels than red or blue sensitive ones, so resolution and sharpness in the green channel will, as a rule, be the highest. Focusing is calibrated for green. When the image is less sharp in the red or blue channels than it is in the green channel, is that because of a wavelength-dependent focus shift or because there are only half as many red or blue sensitive pixels?

 

 

True Michael, but with a home made DNG converter, we can use half of fhe green pixels so we can compare at equal resolution for all colour channels.

[mjh]

True Michael, but with a home made DNG converter, we can use half of fhe green pixels so we can compare at equal resolution for all colour channels.

And have you?

[Lindolfi]

Yes, I have, but only used it on normal images. Still need to test on a Siemens star to see what happens with resolution.

[Lindolfi]

Here's the MatLab code to read the bayer array of an M9 DNG uncompressed file. So you can play with it yourself. It is fun to see what the pixels actually contain before the colour image is composed out of it. By selecting only the pixels that contain red, blue and green you can obtain the three colours and by leaving out half of the green pixels you can have equal resolution in terms of the sensor.

 

%script to read an uncompressed M9 DNG file

close all;
clear;
fname='/Applications/MATLAB74/L1017944.DNG';
fid=fopen(fname,'r');
linelength=10432;

extrapart=(linelength/2);
theoffset=-354-62;
imstart=34400+extrapart*14+theoffset;
numlines=3472;
theheader=fread(fid,imstart,'uint16');
bayerarray=zeros(numlines,linelength/2,'uint16');

for i=1:numlines-1
   bayerarray(i,=fread(fid,linelength/2,'uint16');
end;
fclose(fid);
image(bayerarray);

[brianv]

Looking at the Infrared Focus Mark on several lenses, DigiLLoyd has confused Red with Infrared. There is a big difference in dispersion between the two.

 

Either that, or DigiLloyd has confused "color filters" with Spiratone prism filters.

[Lindolfi]

With the homemade raw dng converter I posted earlier in this thread, I have separated the three partial images from de red pixels, half of the green pixels and the blue pixels. Plotted them and assigned gray values to them. Here are the results from a Summilux 75/1.4 at f2/8 , which is the lens with the strongest longitudinal chromatic aberration in my Leica M collection. The images from an M9 are at full pixel size. As you can see there is no difference in sharpness, so this does not support the observation that color filtering does anything to the focal plane in the case of this lens.

 

http://www.photoplaza.nl/lindolfi/redimage.png

http://www.photoplaza.nl/lindolfi/greenimage.png

http://www.photoplaza.nl/lindolfi/blueimage.png

[MarkP]

Advertisement (gone after registration)

Looking at the Infrared Focus Mark on several lenses, DigiLLoyd has confused Red with Infrared. There is a big difference in dispersion between the two.

 

Either that, or DigiLloyd has confused "color filters" with Spiratone prism filters.

 

Maybe DigiLloyd is just confused.

[luigi bertolotti]

Here's the MatLab code to read the bayer array of an M9 DNG uncompressed file. So you can play with it yourself. It is fun to see what the pixels actually contain before the colour image is composed out of it. By selecting only the pixels that contain red, blue and green you can obtain the three colours and by leaving out half of the green pixels you can have equal resolution in terms of the sensor.

 

%script to read an uncompressed M9 DNG file

close all;
clear;
fname='/Applications/MATLAB74/L1017944.DNG';
fid=fopen(fname,'r');
linelength=10432;

extrapart=(linelength/2);
theoffset=-354-62;
imstart=34400+extrapart*14+theoffset;
numlines=3472;
theheader=fread(fid,imstart,'uint16');
bayerarray=zeros(numlines,linelength/2,'uint16');

for i=1:numlines-1
   bayerarray(i,=fread(fid,linelength/2,'uint16');
end;
fclose(fid);
image(bayerarray);

 

FINE !!! I wonder... a someway similar code could be written to read the NO-Bayer array of THREE MM DNG files... each one taken on the same (static...) scene with three calibrated R,G,B filters... and then to recompose from them a RGB image with a HIGH color depth (14x3=42 bits ?)... a funny academic exercise of "ultra-HDR"...

[chrism]

I have yellow, orange, red and green filters for my most used lenses. I can't say I have noticed any focussing problems unattributable to my aging eyes. Much as I dislike a 'Huffite' adjectival phrase, I don't think this can be a 'real world' problem.

 

Yellow filter: Lucy | Flickr - Photo Sharing!

 

Orange filter: Thoughtful | Flickr - Photo Sharing!

 

Chris

[AlainLandry]

Thank you Jeff! You had answered some of my questions!

 

Alain

 

I've had my MM for about six weeks now. I've taken something on the order of 2400 shots with it. Of those, probably 50-60% used a medium-yellow filter, a smaller number used a yellow-orange filter, and a smaller number yet used a red filter. Lenses have included the Noctilux (no colored filters used), 35 Summilux ASPH (a non-APO lens), 50 Summicron (a non-APO lens), and a 50 Summilux ASPH (a purported APO lens).

 

Empirically - I do photography, not camera-focus or lens-resolution tests - I've noticed no focus issues when using color filters. My focus hit rate has remained the same whether I'm using filters or not. It also approximates what I have long gotten when using my M9 - a camera for which I've never used a color filter. So for me, I can see no practical difference.

 

I'm not disputing the theoretical issue. Quite the contrary, the optical phenomenon that Lloyd alludes to has been known for decades. And yet I don't recollect there was all that much angst around the use of color filters back in the day when we all carried film cameras.

 

It's also possible that if I shot lots with red filters I might start to notice the issue more. But as has already been mentioned, a red filter creates fairly dramatic and sometimes surreal effects - something most of us aren't inclined to use in an everyday, all-the-time fashion. Yellow filters, historically those most used in black and white photography, have little-to-no theoretical focus shift. Certainly not any on a pragmatic basis.

 

Based upon what I'm seeing, focus shift worries when using colored filters on the MM is a bit overblown.

[Lindolfi]

FINE !!! I wonder... a someway similar code could be written to read the NO-Bayer array of THREE MM DNG files... each one taken on the same (static...) scene with three calibrated R,G,B filters... and then to recompose from them a RGB image with a HIGH color depth (14x3=42 bits ?)... a funny academic exercise of "ultra-HDR"...

 

Yes Luigi, that is possible. The three exposures should be perfectly registered (no camera motion down to less than a pixel) and nothing should move in the scene. Could be fun and like most fun things of little practical value

[wattsy]

Yes Luigi, that is possible. The three exposures should be perfectly registered (no camera motion down to less than a pixel) and nothing should move in the scene. Could be fun and like most fun things of little practical value

 

The registration part is the hardest but, in practice, you will not notice (for final reproduction purposes) slight misregistration of a pixel or so. I disagree that it is of little practical value – I quite like the old fashioned (early colour photography) look you get and the slightly spooky effect where parts of the scene move between exposures.

 

If you get the registration right, the result is also very sharp. The following are 100% crops from MM colour test shots I did earlier in the week.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Lindolfi]

Thanks Wattsy, very nice! This can also be done with an M9 or any other RGB camera. And you don't even need the filters in that case, just separate the three colour layers.

 

I agree that it is interesting to first create a monochrom camera and from there try to make colour images again. Many slow forms of photography have their charm and produce their own type of images with interesting defects that can be used in the mood of the image.

[wattsy]

This can also be done with an M9 or any other RGB camera. And you don't even need the filters in that case, just separate the three colour layers.

 

Not really the same thing though is it?

[mjh]

Not really the same thing though is it?

Not at all the same. The only comparison would be to Hasselblad’s multi-shot backs where the sensor can be shifted in both dimensions in pixel-pitch steps. Four shots deliver a full set of RGB data at the nominal sensor resolution, rendering the demosaicing step superfluous.

[Lindolfi]

Not really the same thing though is it?

 

Hence my second remark of the posting from which you quoted.

 

In the light of a photographic technique it is not the same, but in the light of the subject of this thread (let's not forget) it serves answering the same question: "does color filtering affect the focal plane?"

[mjh]

it serves answering the same question: "does color filtering affect the focal plane?"

And the answer is: yes, it does. The amount of the focus shift varies with the filter colour and I wouldn’t worry about yellow or green filters, but when using orange or red filters with the M Monochrom, checking the focus would be advisable. (Admittedly you need a camera with the resolution of the M Monochrom to clearly see the effect; if you are comparing synthesised images derived from an 18 MP CFA image, what you see is how the focus shift would affect a hypothetical 4.5 MP monochrome camera.)

[Lindolfi]

... but when using orange or red filters with the M Monochrom, checking the focus would be advisable

 

Please provide data that show that the effect of focus shift of modern Leica M lenses in terms of a drop in resolution of the analogue image due to filtering (in the range of the visual spectrum) produces an image that has a resolution lower than the sensor of the M Monochrom can resolve, while without filtering the analogue image has a higher frequency content than the cut-off frequency of that sensor.

[mjh]

Please provide data that show that the effect of focus shift of modern Leica M lenses in terms of a drop in resolution of the analogue image due to filtering (in the range of the visual spectrum) produces an image that has a resolution lower than the sensor of the M Monochrom can resolve, while without filtering the analogue image has a higher frequency content than the cut-off frequency of that sensor.

I will deal with that issue (and others) in an upcoming article for LFI. But it is easy enough to check für yourself, provided you have an M Monochrom and an assortment of filters.

[Advertisement]

Hallo,

Look here: Leica M

[Lindolfi]

Looking forward to that article. Especially the data.

Well, I just built a different setup: a microscope with a micrometer looking at the aerial image from a Leica lens while filtered with a green or red filter and measuring the displacement of the focal point with the micrometer adjustment of the microscope. The defect of focus can be calculated into a drop in resolution.