JPEG file size = lens sharpness and/or focus accuracy (?)

[ponch99]

I did test shots of an ISO resolution chart with some lenses and observed that the JPEG file size was very well correlated with the image sharpness (for a lens at different apertures and between lenses).

I think I can almost use the file size as a benchmark for evaluating lens sharpness (and focus accuracy)
What is your experience / what do you think ?

*******************************************
350D, JPEG high, ISO 200, central point autofocus
tripod, light controlled.
Target size: A2, distance depending on FL
*******************************************
Lenses:

• 10-22
• 18-55 II (mine is sharp wide open - center and corners - at all FL!)
• 17-55 IS (two samples, A and B)
• 28-70 3.5-4.5 II
• 50 1.8 II (front focus)
• 50 1.4 (accurate focus)
• 60 2.8

Two shots for each (labelled A and B) at various apertures and FL

Results (name, then size sorted. Size is Ko):
**********************************

 

[Daniella68313]

I think it's locig, since the sharpest images wiill have momre fine detail, that should raise the image size, just like high ISO image are bigger, but that's undesirable detail there.

I did test shots of an ISO resolution chart with some lenses and
observed that the JPEG file size was very well correlated with the
image sharpness (for a lens at different apertures and between
lenses).
I think I can almost use the file size as a benchmark for
evaluating lens sharpness (and focus accuracy)
What is your experience / what do you think ?

*******************************************
350D, JPEG high, ISO 200, central point autofocus
tripod, light controlled.
Target size: A2, distance depending on FL
*******************************************
Lenses:

• 10-22
• 18-55 II (mine is sharp wide open - center and corners - at all FL!)
• 17-55 IS (two samples, A and B)
• 28-70 3.5-4.5 II
• 50 1.8 II (front focus)
• 50 1.4 (accurate focus)
• 60 2.8

Two shots for each (labelled A and B) at various apertures and FL

Results (name, then size sorted. Size is Ko):
**********************************

--



http://www.pbase.com/zylen

 

[ponch99]

http://www.sinepatterns.com/images/ISO-Res.JPG

 

[skyliner]

Wow, that's a lot of data!

It makes sense that this would be the case, because the size of a compressed image file depends on the amount of detail in the image. The same is true of RAW images, which use lossless compression.

Using large apertures is a way to get more photos onto a memory card! Less depth-of-field and lower ISO means smaller files. ISO plays a pretty big role in determining file size as well. Notice that the camera’s estimate of the number of pictures remaining actually decreases as the ISO is increased.

 

[ponch99]

Thanks for reply

I know that the change of file size is logic with jpeg file compression.
The sense of my remark is:

with the same camera (and same in-camera adjustments) is it possible to compare quantitatively image sharpness between different users ?

Example:

I think I have a sharp kit lens (at 3.5-3.8Mb at all apertures and FL), and over 4Mb a lens is excellent.

My 60 2.8 doesn't "go" over 3.7Mb. If other users have 4Mb and more, mine is not good, ...

 

[Glenn W.]

Surely some of this would be effected by the lighting to some degree... and also differnt cameras would yeild differnt results, etc... Whle it may very well be a reasonable way for you to determin your best lenses... I highly doubt that you could ever use it as a 'uinviersal' indicator of lens quality... that is I don't think it would be reasonable for more then one person/camera to compare data...

Thanks for reply

I know that the change of file size is logic with jpeg file
compression.
The sense of my remark is:
with the same camera (and same in-camera adjustments) is it
possible to compare quantitatively image sharpness between
different users ?

Example:
I think I have a sharp kit lens (at 3.5-3.8Mb at all apertures and
FL), and over 4Mb a lens is excellent.
My 60 2.8 doesn't "go" over 3.7Mb. If other users have 4Mb and
more, mine is not good, ...

 

[mdg]

The sense of my remark is:
with the same camera (and same in-camera adjustments) is it
possible to compare quantitatively image sharpness between
different users ?

Example:
I think I have a sharp kit lens (at 3.5-3.8Mb at all apertures and
FL), and over 4Mb a lens is excellent.
My 60 2.8 doesn't "go" over 3.7Mb. If other users have 4Mb and
more, mine is not good, ...

I guess looking at the peak file size from all your images would be a reasonable indicator if you had a lot images, though if you tend to shoot lots outdoors you might wind up with smaller images if you have large ammounts of plain sky, for example.

Noise contributes to jpeg file size so you have make sure your largest image was at a low ISO...

 

[skyliner]

The sense of my remark is:
with the same camera (and same in-camera adjustments) is it
possible to compare quantitatively image sharpness between
different users ?

You might be able to use it as a bechmark for your own lenses under really controlled conditions, but comparing the results cross-user would introduce too many uncontrolled variables. The lighting and exposure will affect it, too. If you're getting a 3MB file and someone else is getting a 4MB file with the same camera, lens, and test chart, that definitely means something if the settings are the same and the faming is basically identical. But, slight variations in the conditions could produce a difference on the order of perhaps as large as 100KB even with equal lens sharpness.

For instance, the futher the color balance is deviated from pure white, the more noise is introduced in the white balancing process. So, if one person tests under good studio light, the file might end up being smaller (due to less noise) than one taken with an equally sharp lens under poor light (but proper exposure and white balance).

 

[Daniella68313]

Wow, that's a lot of data!

It makes sense that this would be the case, because the size of a
compressed image file depends on the amount of detail in the image.
The same is true of RAW images, which use lossless compression.

Using large apertures is a way to get more photos onto a memory
card! Less depth-of-field and lower ISO means smaller files.

lower ISO yes, but less DOF no. he smaller the DOF the lesser the detail since the detail is only concentrated in a small area. put everytyhing in focus and you have detail from edge to edge.

ISO

plays a pretty big role in determining file size as well. Notice
that the camera’s estimate of the number of pictures remaining
actually decreases as the ISO is increased.

--



http://www.pbase.com/zylen

 

[Daniella68313]

Thanks for reply

I know that the change of file size is logic with jpeg file
compression.
The sense of my remark is:
with the same camera (and same in-camera adjustments) is it
possible to compare quantitatively image sharpness between
different users ?

of course not. subject, light, focusing distance etc, would have to be exactly the same.

Example:
I think I have a sharp kit lens (at 3.5-3.8Mb at all apertures and
FL), and over 4Mb a lens is excellent.
My 60 2.8 doesn't "go" over 3.7Mb. If other users have 4Mb and
more, mine is not good, ...

no, you cannot really apply this to it, but if you took 50 shots of the same subject, a good bet for the sharpest pic of it would probably be to look for the larger files.

although I don't think it would be that significant unless some are really soft.

--



http://www.pbase.com/zylen

 

[PeaceFrog]

This is most certainly how I determine the best shot when doing lens tests. I normally take 20 or so shots with a lens at the exact same settings and pick the best one for lens tests. I have found that the file size is a very good way to quickly pick the best of the bunch.

Greg

--



http://www.pbase.com/dadas115/

 

[ponch99]

Of course, you have understood that my test was done in controlled conditions:

• target: resolution chart, identical framing for every lenses
• light: "controlled", so same light conditions
• in these conditions, the focusing distance doesn't seem to play any role

In practice, if the focus is ok, there is only a small file size difference for a given lens at a given aperture with multiple shots.

For me the test is "reproductible", maybe can it be generalized to others

Thanks for reply

I know that the change of file size is logic with jpeg file
compression.
The sense of my remark is:
with the same camera (and same in-camera adjustments) is it
possible to compare quantitatively image sharpness between
different users ?

of course not. subject, light, focusing distance etc, would have
to be exactly the same.

Example:
I think I have a sharp kit lens (at 3.5-3.8Mb at all apertures and
FL), and over 4Mb a lens is excellent.
My 60 2.8 doesn't "go" over 3.7Mb. If other users have 4Mb and
more, mine is not good, ...

no, you cannot really apply this to it, but if you took 50 shots of
the same subject, a good bet for the sharpest pic of it would
probably be to look for the larger files.

although I don't think it would be that significant unless some are
really soft.

--



http://www.pbase.com/zylen

 

[skyliner]

Using large apertures is a way to get more photos onto a memory
card! Less depth-of-field and lower ISO means smaller files.

lower ISO yes, but less DOF no. he smaller the DOF the lesser the
detail since the detail is only concentrated in a small area. put
everytyhing in focus and you have detail from edge to edge.

But if there is less detail, the file is smaller. So why do you not agree that a larger aperture with less of the image in focus will produce a smaller file?

 

[GaborSch]

The JPEG file size is not a good indicator of image quality. A large file can be resulted from many fine details, but noise can cause an even larger file.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

 

[Rick Knepper]

 

[Daniella68313]

Of course, you have understood that my test was done in controlled
conditions:

• target: resolution chart, identical framing for every lenses
• light: "controlled", so same light conditions
• in these conditions, the focusing distance doesn't seem to play

any role

I meant subject distance, not focusing distance, sorry.

-



http://www.pbase.com/zylen

 

[Daniella68313]

Using large apertures is a way to get more photos onto a memory
card! Less depth-of-field and lower ISO means smaller files.

lower ISO yes, but less DOF no. he smaller the DOF the lesser the
detail since the detail is only concentrated in a small area. put
everytyhing in focus and you have detail from edge to edge.

But if there is less detail, the file is smaller. So why do you not
agree that a larger aperture with less of the image in focus will
produce a smaller file?

oh gee..can't beleive I wrote that! you're right of course..I misread what you wrote at first.

--



http://www.pbase.com/zylen

 

[Luna9]

With less information (ie. blurrier), it's able to run more efficient compression.

In the video world, we like to use uncompressed files with a codec like QT Animation. You get huge file sizes, but if you add just a tiny bit of blur, you get a big savings in file size. You can try the same thing with an image in PhotoShop. Draw a circle with 0 feathering. Dupe it and then save it out. Make a dupe of the circle again and add a tiny** bit of blur to it asnd save it out again. You may hardly notice the blur but you'll see it's a smaller file size if you rendered it out with the same level of compression as the unblurred file.