Best file format for viewing images on web?

[george013]

George,
Thanks for speaking out.
Keep in mind that the image is just a matrix of numbers. Therefore we can view the matrix in many ways not just as rasters. We can also prepare the matrix so that it fits various plotting schemes.
I hope to find what we need to understand the image both at different levels and at different steps in its history. You have any ideas? Everyone tells why we can't but I want to know why and can we dream away to understand it.

I don't smoke or drink strange stuff! I was born this way.

RONC

Not just a matrix of numbers. The control the output of your monitor, printer. They control the devices that make the image visible.

I did post this diagram several times and will do it again. Until now now response.

Everything you're doing is done on that raster image. The files you mentioned are disk files, containing the raster image in some way for some purpose. If you want to do pixel peeping you have to find a disk file that saves the raster image without modification. As far as I know that's the TIFF, either 8 or 16 bit.

George

[Rechmbrs]

Manfred,
Have you checked the color type on your facial paint? sRGB or Adobe?
I've covered your comments in other answers except display hardware. Right now I'm doing everything on a laptop with a small screen. Having something bigger like a 4k x 3k at 16bits would be a waste. So I want ways to view the 36mp by 14 bits with what I have.
You going to help figure out how?
RONC

[Rechmbrs]

George,
The raster image is only for the printer and monitor. The tiff, pdf etc are just numbers in a matrix as I said. They have been modified by various algorithm that are applied but they are not raster images. The raster box should be pushed down one level and the matrix of numbers box inserted connected to disc and editor. The editor also links to the rasterizer if you view the data. Really is a different rasterizer than the printer one.

Raster seems to mean different things to different people but the first rasterizers took the numbers and made them suitable for plotting devices like Calcomp, Gould, others. Early monitor output was rasterized by the program. I believe Adobe caused the problem because they called reformatting of the data rasterizing.
RONC

RONC

[Rechmbrs]

I agree on what you say for Flickr etc. But If I load images to CIC to show a point or problem, I can't do it and be happy you see the differences I see. I spend a lot of time helping fix image problems and many are invisible when they post the data. So it's Dropbox the file to me.......
I would like to define a methodology for at least we can view images and see their differences.
RONC

[Mike Buckley]

But If I load images to CIC to show a point or problem, I can't do it and be happy you see the differences I see...I would like to define a methodology for at least we can view images and see their differences.

I don't think that's practical on a widespread basis. You may be able to do it with your next-door neighbor if you go into his home, ensure that he is properly profiling and calibrating the monitor, that he is using the same color space as you, and when using the Internet that he is also using the same browser as you. Otherwise, lacking such precise control over your equipment and other viewers' equipment, there is simply no way to ensure that you and the other viewers are going to see the same thing.

[Manfred M]

Manfred,
Have you checked the color type on your facial paint? sRGB or Adobe?

The image of the facial paint you are seeing will be sRGB, as that is how I posted it to the internet. When I view it on my computer screen, it will be using the AdobeRGB colour space. That's because I have imported it and processed the raw data in the ProPhoto colour space, but my computer uses a 10-bit video card and my display is AdobeRGB compliant. My screen has been calibrated and profiled. I am likely seeing colours more accurately than most people can (I do have to make some allowances for my personal "hardware", as colour vision does decline with age).

I've covered your comments in other answers except display hardware. Right now I'm doing everything on a laptop with a small screen. Having something bigger like a 4k x 3k at 16bits would be a waste. So I want ways to view the 36mp by 14 bits with what I have.
You going to help figure out how?
RONC

I have no idea as to which laptop you are using, but in general most laptops have, at best, sRGB screens with 8-bit video cards. There are a few high end laptops that have AdobeRGB compliant screens and display hardware, but I have never personally run across anyone who owns one. Laptops screens are primarily designed to conserve power (to maximize battery life) and to be as light weight as possible. A lot of the low to medium end ones use TN (Twisted Nemat) technology and it can be argued that they are not truly sRGB compliant as they use a 6+2 bit approach; i.e. they are native 6-bit plus the other 2 bits are emulated through dithering. Higher end laptops (like the entire MacBook line) use true 8-bit displays and are sRGB compliant. This means that there is no possible way for you to display 14-bit colour on these screens on your laptop.

Any image, regardless of the colour space you use, will display out of gamut (OOG) colours based on the rendering intent you have set up in your display settings. For photography, either relative colormetric or perceptual intents tend to be used. The relative colormetric intent will assign the maximim value (255 for 8-bit or 1023 for 10-bit) for any OOG colours. All other colours will be displayed "correctly". The perceptual intent will redistribute all the colours proportionately to ensure all are in gamut. Different suppliers handle this differently, so two different screens (or printers) will handle the colours differently. Most photographers will tend to use the relative colormetric rendering intent, especially when shooting skin tones. Images with vibrant colours can look better when using the perceptual intent.

I've discussed the impact of downsampling a 36MP image for display on what is essentially a 2MP display earlier. You can display parts of your image at 100% in some software, but the whole image will have to be downsampled for it to be seen on a computer screen.

Bottom line is that there is no possible way to display the 14-bit, 36MB image "natively" on your laptop screen. That is a limitation of the hardware.

Finally, most computer screens are set too bright straight out of the box and they need to be calibrated and profiled to give you accurate colours. Screen brightness should be in the range of 80 - 120 candela per square meter (a.k.a. nits) and viewing should be in a dimly lit area (below 70 lux).

[xpatUSA]

Any image, regardless of the colour space you use, will display out of gamut (OOG) colours based on the rendering intent you have set up in your display settings. For photography, either relative [colorimetric] or perceptual intents tend to be used. The relative [colorimetric] intent will assign the maximim value (255 for 8-bit or 1023 for 10-bit) for any OOG colours. All other colours will be displayed "correctly". The perceptual intent will redistribute all the colours proportionately to ensure all are in gamut. Different suppliers handle this differently, so two different screens (or printers) will handle the colours differently. Most photographers will tend to use the relative [colorimetric] rendering intent, especially when shooting skin tones. Images with vibrant colours can look better when using the perceptual intent.

And perhaps we should mention that, for screen viewing, 'perceptual intent' does not work for ICC profiles of the 'display' type, because they are usually of the simple matrix type - with no CLUTs. So, even though the 'display' type of profile usually says 'perceptual', what you get is relative colorimetric.

[george013]

George,
The raster image is only for the printer and monitor. The tiff, pdf etc are just numbers in a matrix as I said. They have been modified by various algorithm that are applied but they are not raster images. The raster box should be pushed down one level and the matrix of numbers box inserted connected to disc and editor. The editor also links to the rasterizer if you view the data. Really is a different rasterizer than the printer one.

Raster seems to mean different things to different people but the first rasterizers took the numbers and made them suitable for plotting devices like Calcomp, Gould, others. Early monitor output was rasterized by the program. I believe Adobe caused the problem because they called reformatting of the data rasterizing.
RONC

RONC

I hope we don't start a discussion about the different meanings of the word raster. If you want we can call it a RGB-raster for the memory file and a R,G,B-raster for the RAW-file. Note the separators. They are not only for printer and monitor. TheRGB-raster files are for editing too. Beside that, what do you mean with "best file format for viewing images on the web"?????
Image disk files are just examples of that RGB-raster file in memory, mostly compressed in some way to save disk size and transferring time.
In that RGB-raster image I can address any pixel with x,y. I can't do that with a JPG or any other diskfile. With the exception of TIFF as far as I know.

As explained also by others here, what you see is what your hardware is capable for. And that's no 14 bits. Go to the properties of your screen. It might be 24 or 32 bits. 24 means 3x8, 32 means 3x8 and 1x8 for some other reason or just using dwords. I don't know.
Anyone who's stating he can see the difference between a JPG and a 14-bit RAW file is telling nonsense. He doesn't know what he compares.

George

[xpatUSA]

I did post this diagram several times and will do it again.



Until now now response.

George

Here's a response, George:

Not all cameras use a color filter array. It should make that clear.

"converter" should say "raw to RGB converter"

"Jpg" should be "JPEG", "Tiff" should be "TIFF".

"editer" should be "editor".

no such word as "scherm" in English.

Inconsistent use of capital letters.

Pardon my pedantry.

[george013]

Here's a response, George:

Not all cameras use a color filter array. It should make that clear.

"converter" should say "raw to RGB converter"

"Jpg" should be "JPEG", "Tiff" should be "TIFF".

"editer" should be "editor".

no such word as "scherm" in English.

Inconsistent use of capital letters.

Pardon my pedantry.

That took a long time!!!!
But if that's all, than the diagram must be functional and I'm happy.

Back to the thread again.

George

[xpatUSA]

That took a long time!!!!

About one minute, actually!!!!

[Rechmbrs]

I have air in my lungs again so let me re-ask my question but from a different view. In my normal processing flow, I view a number of diagnostics. Most of these are displayed using the image display tools in Photoshop. The diagnostics are images but not photos so the values are not SRGB or any other color space. They created from photos so they are tagged with the color space they were derived from. Examples are 2D Fourier transform, modified Sobel transform, simple spatial filters... None are really sRGB.
How do I apply a different set of colors to the individual diagnostic types. I have a scheme for the colors vs numbers in datasets.

Had to move to photoshop because of photo size and bits per color so old proprietary code wouldn't handle.
RONC

[george013]

I have air in my lungs again so let me re-ask my question but from a different view. In my normal processing flow, I view a number of diagnostics. Most of these are displayed using the image display tools in Photoshop. The diagnostics are images but not photos so the values are not SRGB or any other color space. They created from photos so they are tagged with the color space they were derived from. Examples are 2D Fourier transform, modified Sobel transform, simple spatial filters... None are really sRGB.
How do I apply a different set of colors to the individual diagnostic types. I have a scheme for the colors vs numbers in datasets.

Had to move to photoshop because of photo size and bits per color so old proprietary code wouldn't handle.
RONC

That's a total different question.
I don't know what you're up to, I don't have Photoshop. Be aware you use a certain color space so your output device is showing the colors of your image as your camera saw it. Your output device determines the color space you should use.
That eventual correction of your pixel values can be done in 2 ways: change the pixel values so the image can be send to the output device directly, or say to the output device to interpret the values in a certain way.

I'm not sure I do understand your question. But I miss this in it.

George

[xpatUSA]

I have air in my lungs again so let me re-ask my question but from a different view. In my normal processing flow, I view a number of diagnostics. Most of these are displayed using the image display tools in Photoshop. The diagnostics are images but not photos so the values are not SRGB or any other color space. They created from photos so they are tagged with the color space they were derived from.

If that's what Photoshop does, it should not! Any "diagnostic image" derived from a photo should have the photo's meta-data tags removed.

Examples are 2D Fourier transform, modified Sobel transform, simple spatial filters... None are really sRGB.

Therefore, logic dictates that they should not be so tagged or even captioned thus.

How do I apply a different set of colors to the individual diagnostic types?

You apply a different "set of colors" to the photo and then repeat the diagnostic action.

What is meant by "apply"?

What is a "set of colors"?

I have a scheme for the colors vs numbers in datasets.

Which is?

Had to move to [Photoshop] because of photo size and bits per color so old proprietary code wouldn't handle.

On the other hand, my raw shots rarely exceed 12MB and I remain quite happy with 8-bit FastStone Viewer for quick and dirty work . .

I too review diagnostic images taken from photos using the GIMP, RawTherapee, RawDigger, ImageJ or 'Show Image' - albeit none so fancy as yours.

For example, a favorite sport of mine is to check for over-saturation in flower photos (often other people's!). In the GIMP, it is possible to 'decompose' an RGB color image into separate layers of, for example, Hue, Saturation and Value. Each layer can be saved as a new image. Since such layers are saved as monochrome images - to keep the color-space tag with the saved monochrome image would be incredible stupid of the GIMP - although I've never checked . . so please regard that as an assumption rather than fact.

Also, in RawTherapee, wavelet processing is available and wavelet levels can be show as a review image and saved by means of a screen capture. Screen captures, correctly done, should not have color meta-data, or any I suppose!

I am quite impressed that, apparently, PhotoShop can do Fourier and other exotic transformations.

Like George, I do find it quite hard to follow your terminology (some here might not), so please excuse any misunderstandings.

[Manfred M]

I think George is heading in the right direction with his answer.

Photoshop is primarily a tool that was developed for the photography and publishing industry and deals with "accurate" reproduction on a specific family of devices from computer displays (the various RGB colour spaces), the print industry that uses offset presses (CMYK for colour and dot grain for B&W). Mapping as to how to use it in the application you are looking for, if you stick with Photoshop, is something that has to live within the parameters Photoshop was designed to work in.

If you are using a computer display, depending on the quality, it may not even accurately display the sRGB colour space, which encompasses about 33% of the total colours in human visual range. Many recent lower end and mid-range computer screens should be 100% sRGB compliant. Some high end screens targeting photographers and graphic artists are close to covering the AdobeRGB colour space which covers about 50% of the colours in the human visual range. To go higher than that, you are looking at high end ink jet photo printers that get closer to being able to reproduce most naturally occurring colours that humans can see.

There are other colour spaces in use as well and the ones I am aware of are linked to broadcast signals and the capabilities of projection equipment related to showing feature films in theatres using digital projectors.

That takes us to the concept of "gamut"; which identifies the performance of an actual physical device, often given as a percentage of how much of a colour space that the device can reproduce. When some of the data is "out of gamut", often abbreviated as "OOG", these values have to be handled in some way that allows the device to reproduce these colour values.

That brings us back to your question. It sounds to me that your are looking at output from signal processing equipment and what you are looking for is some way of mapping those outputs to a particular output device. It seems to me that the place to start is understanding the capabilities of the output device you are planning to use and then finding some way of mapping the data you have to this output. Colour space is possibly the least important part of your question.

[xpatUSA]

<>

That brings us back to your question. It sounds to me that you are looking at output from signal processing equipment and what you are looking for is some way of mapping those outputs to a particular output device. It seems to me that the place to start is understanding the capabilities of the output device you are planning to use and then finding some way of mapping the data you have to this output. Colour space is possibly the least important part of your question.

I think that Ron should come back and clarify his last question: "How do I apply a different set of colors to the individual diagnostic types?"

For example, he mentioned as being a "diagnostic type" the '2D Fourier transform' - which has no color and bears no resemblance to the image thus transformed. So, applying "a set of colors" to that "diagnostic type" would be meaningless and difficult, if not impossible.

I fear that first we must fully understand Ron's unusual terminology before truly being able to help him with his last question.