Home

Please support mflenses.com if you need any graphic related work order it from us, click on above banner to order!
SearchSearch MemberlistMemberlist RegisterRegister ProfileProfile Log in to check your private messagesLog in to check your private messages Log inLog in

The amazingly high resolution of a Helios 44M-4
Post new topic   Reply to topic
Home   Manual Focus Lenses
View previous topic :: View next topic  


PostPosted: Mon Apr 14, 2014 6:51 pm    Post subject: Reply with quote
Gerald wrote:


You only could distinguish lines with 2% contrast in Koren's illustration because it is noiseless. It is not a real Picture.
This a crop of my test Picture; the contrast is roughly 7-10% under the mark 26. Note the inevitable noise (I shot ISO 100).



Good point on the noise, but when you take the noise into account strictly speaking it is not clear how MTF x is defined. Btw, how does 7-10% figure was obtained, is it what Imatest spits up?


PostPosted: Mon Apr 14, 2014 7:00 pm    Post subject: Reply with quote
Edited

Last edited by bernhardas on Tue May 10, 2016 7:33 am; edited 1 time in total


PostPosted: Mon Apr 14, 2014 7:35 pm    Post subject: Reply with quote
bernhardas wrote:
I am sorry, but my profession has made me into a professional sceptic.

Frankly the estimate of 10% contrast is in my experience as successful as using your hand to feel when the tap water reached 23 degrees. It is in 95% of estimates wrong.

And isn't there a significant difference in the numerical value of the result whether the cut of is at 2% 7% 10% or 12% ???



Very Happy the 100 year old tradition is that everybody does what he wants!
I am sure there is an ISO norm somewhere defined, but even they sometimes grant you latitude in selective input parameters as long as the process and methodology are adhered to.


Let's get the un abbreviated quote from imatest. (Emphasis by me)
Quote:
.
Traditional “resolution” measurements involve observing an image of a bar pattern (often the USAF 1951 chart), and looking for the highest spatial frequency (in lp/mm) where the bars are visibly distinct. This measurement, also called “vanishing resolution”, corresponds to an MTF of roughly 10-20%. Because this is the spatial frequency where image information disappears— where it isn’t visible, and because it is strongly dependent on observer bias, it’s a poor indicator of image sharpness. (It’s Where the Woozle Wasn’t in the world of Winnie the Pooh.) The USAF chart is also poorly suited for computer analysis because it uses space inefficiently and lacks a low frequency reference.
[/b]


Maybe you knew better and should have explained to the USAF, that created the famous USAF 51 chart, that tests of ultra high definition lenses used in military reconnaissance aircrafts based on traditional resolution measurements were a "poor indication of image sharpness". Maybe the whole industry of television equipments was wrong in using a test system based on parallel lines. Maybe Nikon, Zeiss, Leica , were all wrong when they measured the performance of their lenses by the traditional method. Maybe ...

Returning to the real world, it is understandable that Imatest exaggerates the limitations of traditional test. After all, Imatest sells a computerized test system, and has no interest that people continue testing the lenses by the traditional method. The claim that the in the traditional method "the result "is strongly dependent on the observer bias" is not true! The observer to observer variation is only about 5%. The limitations of the traditional method are others. But this a story for another day ...

Last edited by Gerald on Mon Apr 14, 2014 8:15 pm; edited 1 time in total


PostPosted: Mon Apr 14, 2014 7:55 pm    Post subject: Reply with quote
fermy wrote:

Good point on the noise, but when you take the noise into account strictly speaking it is not clear how MTF x is defined. Btw, how does 7-10% figure was obtained, is it what Imatest spits up?




I estimated the contrast by using Photoshop to measure the levels. The white level of the image is ~ 200, and the black level is ~ 50 (the black region of the image is not shown in the crop).

The level of the black lines is ~ 130, and the level of white lines is ~ 140.

The contrast is calculated as follows:
Contrast = (140-130) / (200-50) = 7%

These values are approximate due to the noise. Another person could find slightly different values.


PostPosted: Tue Apr 15, 2014 1:09 am    Post subject: Reply with quote
I wonder if Russia and China had test charts like these:

http://www.dailymail.co.uk/sciencetech/article-2280445/The-mysterious-barcodes-painted-ground-world-actually-used-calibrate-airborne-cameras.html


PostPosted: Tue Apr 15, 2014 1:26 am    Post subject: Reply with quote
Edited

Last edited by bernhardas on Tue May 10, 2016 7:33 am; edited 1 time in total


PostPosted: Tue Apr 15, 2014 1:40 am    Post subject: Reply with quote
Edited

Last edited by bernhardas on Tue May 10, 2016 7:33 am; edited 1 time in total


PostPosted: Tue Apr 15, 2014 1:43 am    Post subject: Reply with quote
bernhardas wrote:


The lens alone is tested on an optical bench.


I fear you do not know that an optical bench uses optical sensors and auxiliary optics. Conceptually, it is not much different from my setup.


PostPosted: Tue Apr 15, 2014 1:53 am    Post subject: Reply with quote
Quote:



I estimated the contrast by using Photoshop to measure the levels. The white level of the image is ~ 200, and the black level is ~ 50 (the black region of the image is not shown in the crop).

The level of the black lines is ~ 130, and the level of white lines is ~ 140.

The contrast is calculated as follows:
Contrast = (140-130) / (200-50) = 7%

These values are approximate due to the noise. Another person could find slightly different values.


This method is guaranteed to give you exaggerated MTF values A more reasonable approach would be to subtract the noise level from the peak white level. For a strong signal the difference at ISO 100 is probably insignificant, however al low MTF the signal is weak and therefore the noise significantly affects the peak values. So it is quite likely that real MTF at the extinction level in your test is somewhere in 2-5% range as Norman Koren claims.

For the sake of the argument, let's assume that the noise level is 1%. That corresponds to the level of white lines with noise subtracted of 140-0.01*200=138 and MTF value of 8/150=5.3

Last edited by fermy on Tue Apr 15, 2014 2:16 am; edited 1 time in total


PostPosted: Tue Apr 15, 2014 1:55 am    Post subject: Reply with quote
Edited

Last edited by bernhardas on Tue May 10, 2016 7:34 am; edited 1 time in total


PostPosted: Tue Apr 15, 2014 9:06 am    Post subject: Reply with quote
calvin83 wrote:
scsambrook wrote:
I love to see these tests, far beyond my ability to replicate them. The best I can manage is to put up boxes of cereal and dog food and take pictures of them with the camera on a tripod.

Would it make any difference if the test chart had its lines printed in different colours? Would some lenses manage green lines better than blue ones, for instance?

Take a look of this test in dpreview.
http://www.dpreview.com/reviews/sigmadp1/page20.asp

Different lens will responds to different wavelength differently. So, a lens may be better manage green lines while other lens are better in red.


Thanks for that link, Calvin !


PostPosted: Tue Apr 15, 2014 11:49 am    Post subject: Reply with quote
fermy wrote:

For the sake of the argument, let's assume that the noise level is 1%. That corresponds to the level of white lines with noise subtracted of 140-0.01*200=138 and MTF value of 8/150=5.3


You are on the right track, but you should consider that:
1) The noise may be positive or negative, depending on the point of the image.
2) The noise affects both the white and black lines.

The values ​​should be averaged along the white and black lines, not only subtracted from the level of the white line.

Thank you for your interesting remarks! Smile


PostPosted: Tue Apr 15, 2014 8:42 pm    Post subject: Reply with quote
I've had two Helios 44M lenses and neither was particularly notable for any quality. That's why I don't have them anymore. Not that they were bad (well one was not so well made fit and finish wise) but at least here in the U.S. you can pick up a bunch of lenses that perform noticeably better for less money.

Most 50s perform very well in the center, and there is always sample to sample variation present. I think somebody did a comparison on this forum of several Domiplan lenses, and the good ones were very good, while the bad ones were pretty bad. And the Domiplan has a much worse reputation than the Helios.


PostPosted: Wed Apr 16, 2014 12:04 am    Post subject: Reply with quote
Gerald wrote:
fermy wrote:

For the sake of the argument, let's assume that the noise level is 1%. That corresponds to the level of white lines with noise subtracted of 140-0.01*200=138 and MTF value of 8/150=5.3


You are on the right track, but you should consider that:
1) The noise may be positive or negative, depending on the point of the image.
2) The noise affects both the white and black lines.

The values ​​should be averaged along the white and black lines, not only subtracted from the level of the white line.


Well, not really, there is no "negative noise" in photography. Noise can add to the level of signal, but can not subtract from it because photosites are additive devices by their very nature. Furthermore, noise is random, some photosites are affected by it, while others are not. If I understood your methodology correctly, to determine the contrast you sampled the darkest and the brightest values. In that case, the darkest values will have the same level as the darkest values in the absence of noise, while the brightest values will have the noise added. Sure, you can average the values along the darkest and brightest lines and then calculate the difference, that would give you correct contrast values for the additive noise, however I doubt you have done that. Did you?


PostPosted: Wed Apr 16, 2014 1:27 am    Post subject: Reply with quote
fermy wrote:

Well, not really, there is no "negative noise" in photography

Sure there are negative noise in photography!
Noise is a fluctuation around an average. It can be positive or negative. For example, suppose you shoot a perfectly uniform wall and then you count the number of photons captured by each sensor cell. Let's say the numbers are: 958 1032 992 1000 1018 ...

Suppose also that the mean is equal to 1000. The noise of each cell is given by:

958-1000 = -42
1032-1000 = +32
992-1000 = -8
1000-1000 = 0
1018-1000 = +18

The noise of the first cell is negative (-42), the second is positive (+32), and so forth.

If you shoot a new picture, the noise of each cell will be very different because the arrival of photons is a random phenomenon, but the mean will remain the same (1000) if the exposure was not altered.


fermy wrote:

Sure, you can average the values along the darkest and brightest lines and then calculate the difference, that would give you correct contrast values for the additive noise, however I doubt you have done that. Did you?


You always doubt what I do! Laughing
What I did was to move the cursor of Photoshop along the lines and eyeballed the averages. A rigorous procedure would require writing a computer program.


PostPosted: Wed Apr 16, 2014 2:01 am    Post subject: Reply with quote
fermy wrote:






Excuse me, are you a mathematician? Or a physicist? Question Question Question


PostPosted: Wed Apr 16, 2014 2:37 am    Post subject: Reply with quote
Edited

Last edited by bernhardas on Tue May 10, 2016 7:34 am; edited 1 time in total


PostPosted: Wed Apr 16, 2014 1:50 pm    Post subject: Reply with quote
Gerald wrote:

Excuse me, are you a mathematician? Or a physicist? Question Question Question


Mathematician

Coming back to our discussion, noise is not necessarily a fluctuation around average. In photographs noise manifests itself as light spots, never as dark spots, which should tell you that noise' contribution is always positive.


PostPosted: Wed Apr 16, 2014 3:35 pm    Post subject: Reply with quote
Gerald wrote:
fermy wrote:

Good point on the noise, but when you take the noise into account strictly speaking it is not clear how MTF x is defined. Btw, how does 7-10% figure was obtained, is it what Imatest spits up?




I estimated the contrast by using Photoshop to measure the levels. The white level of the image is ~ 200, and the black level is ~ 50 (the black region of the image is not shown in the crop).

The level of the black lines is ~ 130, and the level of white lines is ~ 140.

The contrast is calculated as follows:
Contrast = (140-130) / (200-50) = 7%

These values are approximate due to the noise. Another person could find slightly different values.


It's not extremely significant, but be careful about taking values from photoshop. They are not linear, because gamma is applied in the conversion from raw file to jpeg/tiff. Don't know if MTF is defined on a linear scale though.
If linear scaled photos are needed to measure MTF then software for astrophotography can do this (present RAW files linear)


PostPosted: Wed Apr 16, 2014 8:59 pm    Post subject: Reply with quote
fermy wrote:
Gerald wrote:

Excuse me, are you a mathematician? Or a physicist? Question Question Question


Mathematician

Coming back to our discussion, noise is not necessarily a fluctuation around average. In photographs noise manifests itself as light spots, never as dark spots, which should tell you that noise' contribution is always positive.


I agree with you - noise is positive - but not for the reasons you outline. Have you never seen a picture of a clear blue sky where both light and dark spots can be seen ? These correspond to positive and negative deviations respectively from a mean or expected value. A major source of noise is so called photon shot noise which varies with the square root of the signal. The signal is essentially a photon count so one normally takes the positive square root. I'm not sure what kind of physical situation would be described by a negative signal to noise ratio - for example.

A useful account of Noise in DSLRs is given by Emile Martinec : http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html


PostPosted: Wed Apr 16, 2014 9:41 pm    Post subject: Reply with quote
sichko wrote:
fermy wrote:


Coming back to our discussion, noise is not necessarily a fluctuation around average. In photographs noise manifests itself as light spots, never as dark spots, which should tell you that noise' contribution is always positive.


I agree with you - noise is positive - but not for the reasons you outline. Have you never seen a picture of a clear blue sky where both light and dark spots can be seen ? These correspond to positive and negative deviations respectively from a mean or expected value. A major source of noise is so called photon shot noise which varies with the square root of the signal. The signal is essentially a photon count so one normally takes the positive square root. I'm not sure what kind of physical situation would be described by a negative signal to noise ratio - for example.

A useful account of Noise in DSLRs is given by Emile Martinec : http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html


Well, I am not giving any reasons in the quoted piece, I merely provide the empirical evidence. The reason was given a couple of posts above, it is the fact that sensor is essentially photon counting device, it does not subtract. However, after looking at the link that you've posted, I think my original statement is not entirely accurate either. It appears that read noise, for example, can be negative due to voltage fluctuations in the sensor readout circuitry, which can go either way. So I guess a more precise statement is that the noise is mostly positive.


PostPosted: Thu Apr 17, 2014 11:42 am    Post subject: Reply with quote
sammo wrote:
Gerald wrote:
fermy wrote:

Good point on the noise, but when you take the noise into account strictly speaking it is not clear how MTF x is defined. Btw, how does 7-10% figure was obtained, is it what Imatest spits up?




I estimated the contrast by using Photoshop to measure the levels. The white level of the image is ~ 200, and the black level is ~ 50 (the black region of the image is not shown in the crop).

The level of the black lines is ~ 130, and the level of white lines is ~ 140.

The contrast is calculated as follows:
Contrast = (140-130) / (200-50) = 7%

These values are approximate due to the noise. Another person could find slightly different values.


It's not extremely significant, but be careful about taking values from photoshop. They are not linear, because gamma is applied in the conversion from raw file to jpeg/tiff. Don't know if MTF is defined on a linear scale though.
If linear scaled photos are needed to measure MTF then software for astrophotography can do this (present RAW files linear)


Yes, you're right. Programs like Imatest and similar linearize the data before calculating MTF. That said, the above calculation is for illustration only. In the traditional method, the determination of the limit resolution is based on a purely visual judgment, which does not explicitly use the MTF.


PostPosted: Thu Apr 17, 2014 12:28 pm    Post subject: Reply with quote
sichko wrote:
The signal is essentially a photon count so one normally takes the positive square root. I'm not sure what kind of physical situation would be described by a negative signal to noise ratio - for example.

A useful account of Noise in DSLRs is given by Emile Martinec : http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/index.html

The signal-to-noise ratio, SNR, is defined as:

SNR = (signal power)/(noise power)

SNR is always positive because power is a positive quantity. However, the instantaneous value (or the value at each point, if you're talking about image) of the noise can be negative! This is clear if you read carefully this excerpt from the article you cited, where the author comments on shot noise (highlighted text by me):

"An important characteristic of fluctuations obeying Poisson statistics is that their standard deviation - the typical fluctuation away from the average in the typical count - equal to the square root of the average count is itself. That is , if 10000 photons are collected on average , the typical fluctuation away from this average number of photons will be about 100 - Typically the counts will range from about 9900 to 10100 If instead on average 100 photons are collected , the variation from count to count will be +/-10."

The variation from count to count of +/-10 is the shot noise itself! Although the numbers of collected photons are positive, the fluctuations (the shot noise) can be positive or negative, depending on the moment, or the cell.


PostPosted: Thu Apr 17, 2014 10:09 pm    Post subject: Reply with quote
In general the Russian counterparts to their German brothers and sisters are inferior, but with the Helios 44 that doesn't seem to be the case!


PostPosted: Fri Apr 18, 2014 2:40 am    Post subject: Reply with quote
Edited

Last edited by bernhardas on Tue May 10, 2016 7:34 am; edited 1 time in total
Post new topic   Reply to topic
Home   Manual Focus Lenses
Goto page 1, 2, 3  Next
Page 1 of 3