Sensor Resolution and Printing

FoxT

If you are in doubt I suggest you print an 8x10 crop such that the magnification is the same as if you printed the entire photo at A0. This will give an indication of what you are going to get, and is cheap!

The thing I find hard is this:
A 12 Megapixel camera does indeed have 12 million light sensors...BUT
4 million are red
4 million are green
4 million are blue.

Now a 'print pixel' , or a 'dot' is ( i think ) a dot in a picture that can be any colour you like. So you need 3 camera pixels for 1 'dot' b/c a single camera pixel can only capture 1 colour, whereas a single printer dot ( which is, in turn made up of multiple microscopic dots of different colours) can be any colour you want.



From what I have seen on the web etc , the general requirement for high quality print is 300 dpi.Now , 300 dpi = 300x300 dots per square inch.
Thats 90,000 dots per square inch, or roughly 10 sq inches per million dots.

So a 12 megapixel camera can deliver 4 million dots. That is 40 sq inches, approx.
In a 3:2 format that gives you an excellent 9x6 print, approx.

But then you have dithering, and also the fact that the bigger the print, the further away the viewing distance, and the lower DPI you need....Arrgh.

This is all a long way of asking - are there any standard tables available that will indicate how many megapixels you need for a 'great', 'good', or 'adequate' print quality for a given print size?

PS - lovely photo, well done.

- FoxT

charybdis

FoxT wrote: »

If you are in doubt I suggest you print an 8x10 crop such that the magnification is the same as if you printed the entire photo at A0. This will give an indication of what you are going to get, and is cheap!

The thing I find hard is this:
A 12 Megapixel camera does indeed have 12 million light sensors...BUT
4 million are red
4 million are green
4 million are blue.

Now a 'print pixel' , or a 'dot' is ( i think ) a dot in a picture that can be any colour you like. So you need 3 camera pixels for 1 'dot' b/c a single camera pixel can only capture 1 colour, whereas a single printer dot ( which is, in turn made up of multiple microscopic dots of different colours) can be any colour you want.



From what I have seen on the web etc , the general requirement for high quality print is 300 dpi.Now , 300 dpi = 300x300 dots per square inch.
Thats 90,000 dots per square inch, or roughly 10 sq inches per million dots.

So a 12 megapixel camera can deliver 4 million dots. That is 40 sq inches, approx.
In a 3:2 format that gives you an excellent 9x6 print, approx.

But then you have dithering, and also the fact that the bigger the print, the further away the viewing distance, and the lower DPI you need....Arrgh.

This is all a long way of asking - are there any standard tables available that will indicate how many megapixels you need for a 'great', 'good', or 'adequate' print quality for a given print size?

PS - lovely photo, well done.

- FoxT

You raise an interesting point, but your analysis is a bit off.

Firstly, a 12 megapixel bayer sensor will typically have 3 million red pixels, 6 million green pixels, and 3 million blue pixels, not an equal number of each.

The degree to which a 12 megapixel sensor can resolve more than a 3 megapixel image is debatable, the only way to really test it is to point it at a resolution chart and find an optimal debayering algorithm for the sensor's bayer pattern and the resolution test in question. Of course, this would all have to happen in a controlled, lab-style process, completely unrelated to any real-world practicalities. The reality is that your sensor probably won't objectively resolve it's stated resolution even at the best of times, but in virtually all situations, sensor resolution will not be the limiting factor in the sharpness of your images. A debayered image from your camera almost certainly has a higher potential resolution that the amount of pixels on your sensor divided by four, but you're still not going to need it.

FoxT

quote

You raise an interesting point, but your analysis is a bit off.

Firstly, a 12 megapixel bayer sensor will typically have 3 million red pixels, 6 million green pixels, and 3 million blue pixels, not an equal number of each.


- Yes, I had forgotten about that. You are correct. So a 12 Mp sensor gives 3 million 'dots' instead of 4 million - that will be 33 sq inches instead of 40, which is 6.7" * 4.4", theoretically. ( Not sure what happens if you decide to go b/w...)

The degree to which a 12 megapixel sensor can resolve more than a 3 megapixel image is debatable, the only way to really test it is to point it at a resolution chart and find an optimal debayering algorithm for the sensor's bayer pattern and the resolution test in question. Of course, this would all have to happen in a controlled, lab-style process, completely unrelated to any real-world practicalities.

- You are right again, but it sounds like a lot of trouble & effort.


The reality is that your sensor probably won't objectively resolve it's stated resolution even at the best of times, but in virtually all situations, sensor resolution will not be the limiting factor in the sharpness of your images.

- Interesting point. Why is this the case?

A debayered image from your camera almost certainly has a higher potential resolution that the amount of pixels on your sensor divided by four, but you're still not going to need it.

- hmmm. This is definitely one to discuss over a few pints! Does debayering give you more pixels/ie more information? And if it does, then
1- why cant it be applied recursively (Say I do it 3-4 times will it convert my 3 megadots into say 12 megadots)

2 - why wouldn't I need it?


I know my analysis is not fully accurate, and in fact is quite conservative.
I know that 4-6Mp camera sensor can produce good 10x8 prints, and probably even bigger, which is far in excess of what my earlier note would suggest. But I am still stuck with my question...


are there any standard tables available that will indicate how many megapixels you need for a 'great', 'good', or 'adequate' print quality for a given print size?


FoxT

CabanSail

I have moved this into a thread of it's own. It is a very interesting topic on it's own but was drifting OT from the original inquiry.

charybdis

FoxT wrote: »

- Yes, I had forgotten about that. You are correct. So a 12 Mp sensor gives 3 million 'dots' instead of 4 million - that will be 33 sq inches instead of 40, which is 6.7" * 4.4", theoretically. ( Not sure what happens if you decide to go b/w...)

Well, it doesn't literally produce a quarter resolution image, even in you treat the bayer pixels as subpixels. Again, practically and depending on the debayering algorithm and bayer pattern, most bayer sensors will resolve more than a quarter of their stated resolution. If you convert the image to B&W, you're effectively throwing away all the chrominance information in the image and reducing resolution again, but probably not to a quarter, depending on how the image was debayered. If you had a camera sensor without a bayer array (meaning the sensor was monochrome, capable of doing only B&W) you'd have a theoretical resolution of the sensor's stated resolution, but nobody's prepared to market one of these at the moment.

FoxT wrote: »

- You are right again, but it sounds like a lot of trouble & effort.

It is, and it's probably not worth doing as it won't actually influence what you're doing, you'll just know the technical ceiling of your camera, resolution-wise.

FoxT wrote: »

- Interesting point. Why is this the case?*

There are many factors that influence resolution. Firstly, the only parts of an image that could actually be resolved at the sensor's maximum resolution is the stuff that's in focus, or within the image's depth-of-field. For a 12 megapixel APS-C sensor, even these points will suffer from diffraction at any aperture value smaller than approximately f/8, which means that the lens would have to be out resolving the sensor at aperture values greater than f/8 to even approach that resolution. Even then, small amounts of movement from the camera or subject can slighly degrade resolution, even an SLR's mirrorslap can throw away a small bit of potential resolution.

The reality is that sharpness is more complicated than some would have you believe and, as a goal in itself, it's probably not worth striving for in most cases. In all good photographs, it's really never the outright sharpness that makes them good. If you really want the sharpest possible images with the most resolution, shoot large format film, the larger the better.

FoxT wrote: »

- hmmm. *This is definitely one to discuss over a few pints! Does debayering give you more pixels/ie more information? And if it does, then*
1- why cant it be applied recursively (Say I do it 3-4 times will it convert my 3 megadots into say 12 megadots)*

Debayering doesn't add information to an image, but it can recover some resolution by intelligently interpolating the chroma data from the sensor based on a known bayer pattern. There'd be no point to doing it recursively as a debayering process would have to be preformed on a bayer image and the algorithm would most likely have extracted all the additional resolution it could.

FoxT wrote: »

2 - why wouldn't I need it?

Again, you probably won't approach the ceiling resolution of your images outside of very controlled situations, so the additional resolution granted probably wouldn't be necessary. It's nice to have though.

FoxT wrote: »

I know my analysis is not fully accurate, and in fact is quite *conservative.*
I know that 4-6Mp camera sensor can produce good 10x8 prints, and probably even bigger, which is far in excess of what my earlier note would suggest. But I am still stuck with my question...


are there any standard tables *available that will indicate how many megapixels you need for a 'great', *'good', or 'adequate' print quality for a given print size?

The problem is that it depends. It depends on how the image was processed, debayered, focused, the kind of paper and printing process used, the subject matter, a whole load of things. You can't really apply general rules as there are so many factors and the outcome is largely subjective. A good printer will probably be able to eke out a decent print at most reasonable size/viewing distance combinations from files from any modern DSLR that haven't been thoroughly mangled in post-processing.

stcstc

your right about this geing hard to write a table for

I started to try and found it just waaaay to complicated to explain

on my site in the downloads section is a pdf with print sizes and 3 resolutions, it gives you the number of pixels for each

but all of the theory doesnt matter if you look at some of the prints i have been able to do from some silly small images

for example i recently printed a 15*10 image from a 300k jpg, it looked great. mind the starting image was a good quality shot etc to start with

FoxT

Thanks Charybdis - that was very informative. Prompted by your reply, I have read a few articles on the net about this as well. I think it is fair to say that if I spent more time taking photos & less time worrying about the technology my photography would improve


- FoxT

AnimalRights

I got some photos printed in The Camera Centre last week 8x10s
My files were 25mb a pop so it took a while for the machines to process etc
The guy who worked there said I should reduce the files and DPI of 300 I think he said was ok...is this ok?
I always want the highest quality, I have an important nixer coming up and want to do it right.

CabanSail

300dpi is the industry standard and will be very good quality. Steve uses 360dpi if he can as, if I get this right, it's the native resolution of his printers.

stcstc

you should always aim to use the native resolution of the printers if you can

this way the printer isnt doing any changing of pixels ie interpolation or anything

so for mini lab type stuff 300 ppi (not DPI) is the correct thing to use

so you should output from photoshop or whatever

the correct number of pixels for the print size you need

so 10 inches would be 3000 pixels etc


hope this helps