Picture sharpness and air effects

[Andre]

this 100% crop (format png, 563kb) says something about long distance distortion and sharpness:

meanwhile, another point I missed to make, it seems that the lines and sails get more blurred in the bottom part where they get in the 'ground effect'. Agree?

 

[sophiecentaur]

It's amazing that we see the rigging at all at that distance. It can't be bigger than 10mm and could be 6 or 8mm. (less than 10^-5 radians at a km distance). For an aperture of 20mm and a wavelength of 800nm, the Rayleigh criterion gives a resolution of 2X10^-5. No wonder it's all a bit fuzzy - right near the limit for discriminating two adjacent rigging lines so when they cross, that blur is only to be expected.
Whatever our equipment can do for us, we always want better.

 

[Borek]

Do many cameras actually use lossless JPEG?

I am not aware of any. Some cameras can save in jpg and tiff (that was the case of Marzena's Lumix) - but I never bothered to check the difference between these tiffs and jpg, and I can't don't have access to the camera now. Canon uses either jpg or their own raw format (cr2).

 

[Borek]

this 100% crop (format png, 563kb) says something about long distance distortion and sharpness:

Same crop but now in moderately high quality jpg, size 24kb, notice the minor artificialities?

Small trick - combine the images in the Photoshop, set blending to "Difference", and modify histogram to brighten the image (this one was modified automatically):

Compression artifacts are always best visible close to the edges.

 

[sophiecentaur]

Nice. Note those identifiable 'ripples' on the right of the backstay - they extend way out in the difference picture.

It's even better when this is done with low noise pictures, made under really good conditions. That, of course, is a good way of assessing, objectively, how well a particular compression algorithm is working.

 

[sophiecentaur]

I always wonder why different manufacturers have so many RAW formats /algorithms. Is it Licence issues?

 

[Andy Resnick]

It's amazing that we see the rigging at all at that distance. It can't be bigger than 10mm and could be 6 or 8mm. (less than 10^-5 radians at a km distance). For an aperture of 20mm and a wavelength of 800nm, the Rayleigh criterion gives a resolution of 2X10^-5. No wonder it's all a bit fuzzy - right near the limit for discriminating two adjacent rigging lines so when they cross, that blur is only to be expected.
Whatever our equipment can do for us, we always want better.

That's the difference between 'resolving' an object and 'detecting' the object. It is routine to optically detect particles a few nm in size- either by sidelighting (ultramicroscopy) or direct detection (single-molecule fluorescence). Similarly, we can easily detect stars but we require complicated equipment to resolve them (say, measure their size).

 

[Andre]

Nice Borek

It seems that it sure pays to invest in external drives to save all the raw/CR2's.

About the original picture, Sophiecentaur, to enhance sharpness, noise reduction was turned off and I applied moderate sharpness, 5 out of 10, while Canons standard is 3 out of 10.

hmm I try to understand how Gimp 2.6 works for these kind of tricks but it seems frustratingly different from photoeditor and paint. Would it help to buy photoshop. But there is always the SAF, being rather low in my case.

 

[Andy Resnick]

No, but we shot mostly in RAW, and what we see on our monitors is in RAW as well, we just convert for jpg to show pictures on forum. At this stage - I can speak for myself only, but I doubt And² do it much differently - I choose compression level so that the effect I am referring too is still visible. As far as I know at zero compression level jpg is a loseless format, be sure I will not hesitate to use it if I will find it necessary.

I exclusively shoot JPG. I can see the compression artifacts on occasion, but so infrequently that I can't justify the added time and effort needed to deal with RAW.

 

[Andy Resnick]

<snip>
On the subject of Haze; I have looked all over for a filter that would (sharp-) cut out the far end of the visible Blue. This, I am sure, would reduce the worst of the haze - allowing more contrasty pictures - without knocking out all the Blues. Distant mountains can be made to look a lot sharper if you reduce the gain in the blue but this, of course, wrecks the colour balance. I'm convinced that some clever filter could help a lot.
Why isn't there anything about? (or is there?) It could be that filters (other than polarising) are not used much with digital photography because Photoshop can do most of it after the event.

http://www.tiffen.com/filters.htm and http://www.hoyafilter.com/products/hoya/index.html have some specifically to address haze. Alternatively, you could get a longpass cut-off filter at nearly any wavelength you want:

http://www.edmundoptics.com/products/displayproduct.cfm?productID=2683

 

[sophiecentaur]

I exclusively shoot JPG. I can see the compression artifacts on occasion, but so infrequently that I can't justify the added time and effort needed to deal with RAW.

RAW files are hardly more trouble, actually. They transfer straight into Lightroom or Aperture (there are a lot of alternative 'management' systems on PC, I think, but a good one may cost some money), just taking a bit longer for the files to move. After that, you can see them, sort them and reject some, just as easily as with Jpegs. I used to use Iphoto which didn't deal with Raw (the latest version may) and that will do 'faces' and 'locations', if you're into social media.
For the small number of pictures that you may want to print, email etc., those operations all all integrated. It really is useful to be able to batch process rough colour balance and exposure with the RAW processing - and, on top of that, you have the very desirable exposure latitude that Jpeg lacks. I can really recommend it. As a technical person, you would surely appreciate it. If all you use at the moment is Windows folders or the Camera manufacturer's own software then you could do a lot better. I have seen Pentax, Nikon and Sony's efforts and they are just not very special. Adobe and Apple have certainly got these things sussed. (at a price, of course)

 

[Borek]

One the main reasons for using RAW - for me - is that they are much easier to correct later. Plus they save all 12 bits, so there is a little bit more room for mistakes.

And I agree with sophiecentaur that raws are not more trouble than jpgs are. I browse them using the same programs, I almost never use pictures directly, so I always have to rescale them before posting - and time required to do it is the same, no matter if I start with jpg or raw. If I need to correct histogram, white balance, whatever - using jpgs is not faster. The only drawback is disk space - but storage is not that expensive.

 

[Andy Resnick]

Here's something to consider- these two images (300%) may appear to have different levels of detail:

[PLAIN]http://img821.imageshack.us/img821/7642/dsc0897v.jpg

[PLAIN]http://img9.imageshack.us/img9/7364/83193575.jpg

but linescans across an edge show that the image quality is the same:

[PLAIN]http://img545.imageshack.us/img545/5623/plotofdsc0897.jpg

[PLAIN]http://img502.imageshack.us/img502/1189/plotof07.jpg

One reason is that our eyes are more sensitive to contrast at mid-range spatial frequencies, and so the cutoff frequency does not directly correlate with how 'good' the image 'looks'.

 

[Andy Resnick]

RAW files are hardly more trouble, actually. <snip>

This isn't something I'm going to argue about. They are more trouble for me, so I don't deal with that format.

 

[sophiecentaur]

This isn't something I'm going to argue about. They are more trouble for me, so I don't deal with that format.

Perhaps they are just more trouble in the application you are using. Using Aperture 2, you wouldn't be aware whether you were importing Jpeg or RAW until you wanted to change exposure / brightness/ contrast etc. At that stage you would find that could actually (but only if you wanted to) alter the daylight / shade etc. colour balance, rather than have to frigg it using tone /hue / saturation controls.
When I first used my present camera, I stuck with Jpeg, for the same reason as you - my software made it a real effort. Now it's not, so I use it. You have no idea just how many iffy pictures you can rescue when they're in RAW.

 

[sophiecentaur]

Here's something to consider- these two images (300%) may appear to have different levels of detail:


One reason is that our eyes are more sensitive to contrast at mid-range spatial frequencies, and so the cutoff frequency does not directly correlate with how 'good' the image 'looks'.

From the look of the two graphs, it seems to me that the second picture could have had had some 'aperture correction' processing. The slope is increased and there is a slight ringing before and after. That is a common 'sharpening' method which accentuates edges and makes them more noticeable. In one dimension, it's achieved by a simple delay line filter to introduce a bit of hf boost. Point and shoot digital cameras have this built in, to some extent.

 

[AlephZero]

I am not aware of any. Some cameras can save in jpg and tiff (that was the case of Marzena's Lumix) - but I never bothered to check the difference between these tiffs and jpg, and I can't don't have access to the camera now.

The old "devil's dictionary" definiton of TIFF was "Thousands of Incompatible File Formats". TIFF is really just a container for a whole range of different image formats, some raw, some with lossless compression, some with lossy compresssion. IIRC you can even store "standard " JPG compressed images inside a TIFF. So the .TIFF suffix doesn't tell you much unless you know what is inside any particular file.

I always wonder why different manufacturers have so many RAW formats /algorithms. Is it Licence issues?

I don't have any problem that every manufacturer would have a different algorithm for converting the output from the CCD (which at the level of the electronics starts off as analog data, not digital) into an image file. After all, if they don't know more about the exact properties of their optics and electronic sensors than anybody else, why would you want to buy their cameras?

But the plethora of different image formats for what is conceptually the same data (i.e. a "raw" image, hopefully in some well-defined color space, with enough digital resolution to represent all the information but not so much that the low order bits are just noise) does seems to be rather pointless.

 

[sophiecentaur]

http://www.tiffen.com/filters.htm and http://www.hoyafilter.com/products/hoya/index.html have some specifically to address haze. Alternatively, you could get a longpass cut-off filter at nearly any wavelength you want:

http://www.edmundoptics.com/products/displayproduct.cfm?productID=2683

Thanks for looking but I got that far when I first looked into this. What I want isn't a 'haze' filter, which would have to 'look OK' without any colour correction. What I am after is a filter that cuts out a fair amount of visible blue - but with a sharp characteristic so that there is still a lot of the blue end spectrum still at full amplitude. You'd obviously to have to do some colour correction afterwards and the colourimetry would look odd, no doubt. But the picture would look a good deal sharper because the spectrum of haze drops off significantly over the top (blue) end range. Not something that Hoya etc. are likely to market, I think.
Problem is that the required filter would need to be of a reasonable optical quality or any possible advantage would be undone. A 'bespoke' filter could cost a bomb. I guess a dichroic design could be made. Sods law, actually. Only 40 years ago, the place I worked had a vacuum unit for producing high quality dichroic filters for colour telecine machines. Never in the right place at the right time.

 

[Andy Resnick]

From the look of the two graphs, it seems to me that the second picture could have had had some 'aperture correction' processing. The slope is increased and there is a slight ringing before and after. That is a common 'sharpening' method which accentuates edges and makes them more noticeable. In one dimension, it's achieved by a simple delay line filter to introduce a bit of hf boost. Point and shoot digital cameras have this built in, to some extent.

Both images are straight off the camera (original shot was in B&W), same aperture setting, etc. (manual focus lens). After capture, both images were treated the same- neither was sharpened.

 

[Andre]

I exclusively shoot JPG. I can see the compression artifacts on occasion, but so infrequently that I can't justify the added time and effort needed to deal with RAW.

One the main reasons for using RAW - for me - is that they are much easier to correct later. Plus they save all 12 bits, so there is a little bit more room for mistakes.

..., so I use it. You have no idea just how many iffy pictures you can rescue when they're in RAW.

Maybe I can illustrate the advantage of 12 bits of RAW or more (wasn't it 14?) versus 8 bits JPG as hinted by Borek and Sophie.

Suppose that the metering tells you that this is the perfect exposure setting and that's what you save as jpg:

Obviously the dynamic range here exceeds what 8 bit jpg can reproduce. It won't help much if you increase or decrease the pitch"

You simply lost the information stored in the excess bits, but look what happens if you change the pitch in raw up or down:

See how much information was lost in the jpg.

Now you just need the software to HDR those three together.

 

[sophiecentaur]

Both images are straight off the camera (original shot was in B&W), same aperture setting, etc. (manual focus lens). After capture, both images were treated the same- neither was sharpened.

How do you know what the camera did with its original image if you just trusted its jpeg processing? Those two graphs show that the two pictures differ in more than vertical gain. Consumer cameras do their best to produce a nice picture. They aren't 'measuring instruments'

 

[sophiecentaur]

Maybe I can illustrate the advantage of 12 bits of RAW or more (wasn't it 14?) versus 8 bits JPG as hinted by Borek and Sophie.

Suppose that the metering tells you that this is the perfect exposure setting and that's what you save as jpg:


Obviously the dynamic range here exceeds what 8 bit jpg can reproduce. It won't help much if you increase or decrease the pitch"


You simply lost the information stored in the excess bits, but look what happens if you change the pitch in raw up or down:


See how much information was lost in the jpg.

Now you just need the software to HDR those three together.

It's not for nothing that they sometimes refer to RAW as Digital Negative. You take that 'negative' and 'print it' with various virtual 'grades of paper', you can dodge and burn and a lot of other stuff that will degrade a picture that has already been jpegged.
Raw is the way forward for serious photography. Jpeg is the equivalent of the Instamatic, by comparison.

 

[Andy Resnick]

That's a different problem, one of excessive magnification- even a tack-sharp 35mm image will look horrible if enlarged to a poster-sized print and viewed close-up. That's why medium format cameras are used even for 8" x 10" magazine prints- the magnification is less.

Just to finish this thought- here's the full image field of one of the luminars, projected onto a piece of paper:

[PLAIN]http://img88.imageshack.us/img88/8945/fullimagefield.jpg

For reference, I drew a rectangle on the paper (the small rectangle in the center) the size of a 35mm format image (24 x 36 mm).

Using 300 dpi print resolution, if my camera produces images that can be enlarged up to 20" across, I could enlarge the full image up to 12 feet across and maintain the same level of print quality. In terms of a view camera, using a medium format negative allows a reasonable amount of lens movement to adjust perspective (and film plane adjustment to maintain focus).

 

[Andy Resnick]

Maybe I can illustrate the advantage of 12 bits of RAW or more (wasn't it 14?) versus 8 bits JPG as hinted by Borek and Sophie.

<snip>

Or you could use a density gradient filter in the first place.

 

[sophiecentaur]

And have a special shaped gradient filter for all occasions? I think you need to enter the 21st century, Andy. If you are interested enough to be discussing things at this level you can't justify sticking to jpegs. Pros use RAW for a good reason.

 

[Andy Resnick]

And have a special shaped gradient filter for all occasions? I think you need to enter the 21st century, Andy. If you are interested enough to be discussing things at this level you can't justify sticking to jpegs. Pros use RAW for a good reason.

Until you start posting photos you took, I'm not sure why I should take you seriously.

 

[sophiecentaur]

Until you start posting photos you took, I'm not sure why I should take you seriously.

That's an interesting response. I can't recall, on any previous occasion, being asked, on PF, for personal experimental results, to justify a very reasonable opinion. Did you object to the, perhaps, cheeky wording of my last post. I'm sorry if you were but I thought these conversations were somewhat 'between friends'.

A few minutes on any Photograpy Forum will produce loads of opinions in favour of using Raw format and will provide you with plenty of examples of suitable software. If you have actually used one of the 'modern' image management applications then I would be amazed if you were to say you could tell the difference between the way they deal with Jpeg and Raw formats. If you haven't, then I suggest you give it a try. You will see what I mean.

Can there really be any doubt that data compression before processing cannot produce as good results as processing first and then compressing the data? A 'mechanical' form of pre-processing (a shaped filter mask) would clearly only be a partial solution to the problem of the contrast ratio of an original scene. I don't think you were really being serious, actually.
A few extra bits of quantising must contribute significantly to exposure latitude. Anyone who has used colour reversal film (slides) will know that what you had is what you get. You have nothing like the flexibility that colour negative film will give you. Cibachrome could produce absolutely stunning results - but only from a perfect positive transparency.

http://gallery.me.com/lyner" but not all of high technical quality, of course. Many of the earlier ones were shot in Jpeg.

 

[Andre]

For some more details on the discussion, this may be useful.

Let me just quote the last sentence.

The techniques discussed here work best when full highlight and shadow detail have been captured by the camera sensor in Raw format.

 

[Andy Resnick]

That's an interesting response. I can't recall, on any previous occasion, being asked, on PF, for personal experimental results, to justify a very reasonable opinion. Did you object to the, perhaps, cheeky wording of my last post. I'm sorry if you were but I thought these conversations were somewhat 'between friends'.

A few minutes on any Photograpy Forum will produce loads of opinions in favour of using Raw format and will provide you with plenty of examples of suitable software. If you have actually used one of the 'modern' image management applications then I would be amazed if you were to say you could tell the difference between the way they deal with Jpeg and Raw formats. If you haven't, then I suggest you give it a try. You will see what I mean.

Can there really be any doubt that data compression before processing cannot produce as good results as processing first and then compressing the data? A 'mechanical' form of pre-processing (a shaped filter mask) would clearly only be a partial solution to the problem of the contrast ratio of an original scene. I don't think you were really being serious, actually.
A few extra bits of quantising must contribute significantly to exposure latitude. Anyone who has used colour reversal film (slides) will know that what you had is what you get. You have nothing like the flexibility that colour negative film will give you. Cibachrome could produce absolutely stunning results - but only from a perfect positive transparency.

http://gallery.me.com/lyner" but not all of high technical quality, of course. Many of the earlier ones were shot in Jpeg.

PF is a scientific forum- claims are supported or refuted with evidence, not by appeals to authority or anecdotal evidence.

Again, I have shot 10952 images with my sony (so far), and while 10952 of those images could all be substantially improved, I would be hard-pressed to identify a single image that would be improved solely by a change in file format. As evidence for this, I often post photos here.

None of my scientific images are acquired using the Sony- the Bayer filter, by definition, means that I cannot use those images as quantitative imaging data. I use monochrome focal plane arrays for scientific imaging. Those cameras are calibrated and validated.

You are free to use whatever tools you wish to create an image. Again, and I thought I was being clear about this, in my *personal experience* taking non-science images, I have no use for the RAW data format. *For me personally*, it's not worth the order of magnitude difference in file size or time required to sit and wait for the computer. *Personally*, I prefer to take photos rather than stare at a computer monitor. YMMV.

I don't understand why you are so insistent that I conform to your ideas of what is "best".

 

[Andy Resnick]

For some more details on the discussion, this may be useful.

Let me just quote the last sentence.

A very reasonable article. I focused on the sentence "when seeking the most *realistic* results" (emphasis mine). I honestly don't know what "realistic" means in the context of photography.

Photography is more than simply trying to recreate what you see by eye. If it was, all my images should be blurry and lacking reds and greens :) Photography is also a means to create what I *imagine* I see or to create a scene that is *impossible* to see- here is an example of each:

[PLAIN]http://img27.imageshack.us/img27/2046/8998203090105.jpg

[PLAIN]http://img820.imageshack.us/img820/8479/5828303090105.jpg

In the first image, I wanted to capture the way the moon appears nestled in the clouds like a yellow-red glowing pearl. This is not at all how the scene appeared to my eye- the only thing I could see was the moon in the clouds, the rest of the field appeared black.

The second image could not be seen by eye- the field of view is more than double what our eyes are capable of. Secondly, the sky would be black- I could not see the clouds or horizon.

As a side note for the lurkers, here's a shot my wife took using her point-n-shoot panasonic, all auto setting, etc.:

[PLAIN]http://img33.imageshack.us/img33/757/2572832980105.jpg

The image is unremarkable, to be sure- it's a generic snapshot. Here's a 1:1 crop:

[PLAIN]http://img716.imageshack.us/img716/4518/25728329801051.jpg

The level of detail is good, dynamic range is good, etc. etc. This may not seem to be a big deal, but the camera is a mess- there is sunscreen wiped all over the front lens, the metal lens cover is broken and has to be manually pried open, and is visible in the top-left edge (hence the sunscreen). She didn't any give thought to making the image "look good"- something caught her eye (the transparency of the water) and she pointed and shot. The only processing I did on this image was a contrast adjustment.

My point is that you do not need high-end equipment to produce high quality images- all you need is a clear idea of what you want the photo to look like and some experience learning how to use your camera.