why megapixels keep growing?
Hi all,
I wanted to discuss with you why you think the megapixel count is increasing over the years. As the sensor size remains the same one expect that by just increasing the resolution the captured image would be more noisy.
I had a look on images from older cameras like the nikon d700 that, If I am not wrong it was 12mp camera, and their iso at 3200 or 6400 is much better even compared to todays cameras.
I am shooting with a 16 megapixel camera and I would be happy to even shoot at 12 or 8 Mpixels if there was way to give me less noise per capture.
I wonder though why the trend instead of producing less noisy captures is pushing the number of megapixels to grow. Why I can not have today a low noise 12 Mpixels APS-C or FF frame?
I understand that the megapixel count can put a limit to print sizes, but for my work, 12 or even 8 Mpixels would be more than enough.
I am waiting for your feedback for a nice discussion.
Bring coffee
Alex
I wanted to discuss with you why you think the megapixel count is increasing over the years. As the sensor size remains the same one expect that by just increasing the resolution the captured image would be more noisy.
I had a look on images from older cameras like the nikon d700 that, If I am not wrong it was 12mp camera, and their iso at 3200 or 6400 is much better even compared to todays cameras.
I am shooting with a 16 megapixel camera and I would be happy to even shoot at 12 or 8 Mpixels if there was way to give me less noise per capture.
I wonder though why the trend instead of producing less noisy captures is pushing the number of megapixels to grow. Why I can not have today a low noise 12 Mpixels APS-C or FF frame?
I understand that the megapixel count can put a limit to print sizes, but for my work, 12 or even 8 Mpixels would be more than enough.
I am waiting for your feedback for a nice discussion.
Bring coffee
Alex
0
Comments
2, cropping
3, downsize a 22mp or 36mp image to 8 or 12mp and noise goes away.
Check out the Sony A7s if you really want 12mp FF.
An "accurate" reproduction of a scene and a good photograph are often two different things.
A former sports shooter
Follow me at: https://www.flickr.com/photos/bjurasz/
My Etsy store: https://www.etsy.com/shop/mercphoto?ref=hdr_shop_menu
It's not just process technology but pixel size, spacing, percentage of area taken by control logic, etc. Then of course, the design of the device. Not all of which are equal weight but which must be balanced based on the application.
what I have found is that by increasing the megapixels you increase resolution at base iso. It is a trade off, if you reduce pixel count resolution at low isos falls but increases (compared with a higher megapixel camera) in high isos.
The questions though is
-How much megapixels you need? This somehoe relates with print sizes. So to simplify further the question what print sizes (A4,A3,A2) a 6Mpixel image can give? a 12?16?24? and so on. If I know that not any customer would ever ask for sizes up to A3 it would be nice to have the options (through market products) to pick the megapixel count that can cover your print sizes you want and trade the rest with improved low light quality
-I do nopt have a d700 or d800. I looked on dxomarks some time ago but to be honest, even though I am coming from an electrical engineering background, I do not care much for the numerology. Every time I look a shot is not the bad iso performace for sure (with minor exceptions). I do not think that photographers one decade ago they were filling that 12Mpixels were not enough.
Thanks for sharing your thoughts
For instance, assuming a Bayer array sensor design:
Each "sensel" (Sensor Element) in a Bayer array imager only represents one of three colors; Red, Green or Blue. Additionally, most of the luminance information comes from the Green channel. There are twice as many Green sensels as there are either Blue or Red sensels.
All of this means that for a Bayer array sensor digital camera, each of the "pixels" of an image is interpreted from other, surrounding pixels. The "actual" luminance information, where the image details come from, are around half of the manufacturer's rated "pixel" rating (counting only the Green sensels.)
Similarly, the Green channel has twice the sensel data of either the Red or Blue channels, meaning that image portions are more accurately color represented by color tones and hues with mostly green information. Imagers must also contend with color sensitivity, meaning that some color tones are more faithfully represented than other color tones, and some color tones will have more color noise than other color tones, again mostly due to differing color sensitivity of each sensel.
Basically, all of the above means that if you properly resample "any" Bayer imager to 1/4th of the manufacturer's rated resolution, you should have an image with "more" accurate color and detail information than the Bayer alone representation resulting from most 1:1 Bayer demosaicing algorithms (which generally calculate 1 sensel equaling 1 pixel).
Since all images are interpolated for both luminance and color information, I regularly interpolate to a larger resolution as a better starting point for processing images. This supersampling should equate to better overall image rendering for enlargements after post-processing, especially if they are downsampled to less than rated resolutions for output. (Note that photo-finishers do their own interpolation and resampling as part of the printing RIP Engine (Raster Image Processing Engine), which generally means you don't have to do that step for printing images on their equipment.)
Smoother anti-aliasing of sharp edges and better control of USM sharpening halos are two tangible benefits of super-sampling RAW images in the RAW processor work stream, for instance.
Moderator of the Cameras and Accessories forums
ISO 3200 on the D700 at 12MP is not even CLOSE to ISO 3200 on the current crop of 24Mp cameras, and even the D810 with 36MP is cleaner at ISO 3200.
In many instances, I agree with you. There are certainly times when I am not "glass limited" and would love to have more ISO performance instead of more megapixels. But, this is not likely to happen.
Both are happening. Cameras are getting less noisy, AND the megapixels are growing. 10 years ago, we spent $6000 to get good ISO 800 performance with 10MP. Today, I can get 24MP, less noise, and spend 1/10 the money. To me, that is excellent progress.
For many, that is not enough MP. And it's not only for print sizes. The ability to crop is EXTREMELY helpful. For the sake of discussion, let's say there was a 100MP camera that could shoot clean ISO 3200 images. For me, that would mean that Instead of buying a 400mm lens for $12k, I could shoot my football, soccer, etc. with an 85mm lens and still get closeups from the other side of the field. That saves me a LOT of money, it means I don't have to carry around a massive amount of weight, it means I don't have to pay extra costs when flying, etc. These are all wonderful things for me.
perroneford@ptfphoto.com
I could not disagree more. But then a LOT of this has to do with your working environment. When you have to live at ISO 3200-12,800 to get your shots like I have to, noise is a constant enemy. When we reach a point that cameras of tomorrow at ISO 3200 look like today's cameras at ISO 100, I'll sing a different tune.
perroneford@ptfphoto.com
Think there's a possibility (for some, maybe?) of losing / forgetting the difference between Mp (area) and focal length / field of view (linear) issues here + shifting datums ... suppose I could say, moving goalposts?
pp
Flickr
perroneford@ptfphoto.com
Wasn't disputing / disagreeing with your points, only that with various Mp numbers being used in the thread, it could be argued that the basis for the 400 / 85mm comparison could get lost.
eg, using slightly different Nos
Say tog currently using a 24Mp body + 400mm lens.(4k x 6k pixels)
Tog just fills the frame with a soccer ball ... therefore it's a 4k x 4k circular area.
Tog then keeps everything (else) the same, but uses their new fangled 96Mp body (8kx 12k) to take a pic from same position.
Ball again fills the frame, but is now an 8k x 8k circular area ... ie double the height / width ... but 4 times the area / number of pixels .
To obtain the same number of pixels on subject with the new cam as with the first, they would have to use a 200mm lens.
So - to my mind, exchanging a 400 for an 85mm (on a 100Mp body) rather depends on what the tog was using to start with for it to make any sense ... 'cos if they were using a 24Mp body they'd still need a 200.
By my maths, an 85mm on a 100Mp body is producing the same No of pixels on target as a 400mm on a 4.5 Mp body (approx)
400 / 85 = 4.7
4.7 x 4.7 = 22.09
100 / 22.09 = 4.53
All assuming my maths is correct, of course ... which is certainly not guaranteed
pp
Flickr
The difference in weight between a 300mm lens and a 400mm lens is ~two pounds of weight and slightly more than US$6000 difference in cost. Only rudimentary financial mathematics is required to see the benefit there!
perroneford@ptfphoto.com
Can't help thinking there's also going to be dof issues too with this approach - especially as the lens focal difference increases?
pp
Flickr
Probably so but that isn't the point. If it is an issue then the photographer would use a 400mm over a 300mm. For 90% of the needs DOF isn't an issue.
-P
perroneford@ptfphoto.com
Well, youve got a couple of really nice, convenient Nos there for tests / experiments etc ... the 2 extremes, 4 and 36 Mp, since 36 is obviously 9x the 4.
therefore ... with the previous caveat that my thinking / maths is correct (haha) ... a shot that fills the frame of the 4Mp would only need to be a third the height /width of the frame on the 36Mp body to have an equivalent No of pixels (imagine the frame chopped up into 3 rows of 3, 4Mp areas).
To achieve this, assuming both bodies of the same crop factor, a lens of a third the focal length used for the 4Mp body has to be used on the 36Mp body.
Cutting to the chase
Take a pic that fills the frame with the 4Mp body + 400mm lens
then, keeping distance the same, use a 70 - 200 @133.33 (400/3) or a 135 (near enuf?) and compare dofs for same apertures?
Another thought
If Mp figures keep climbing ... are current lenses going to be able to cope ... with say,the 100Mp previously mentioned?
pp
Edit
A quick shufties at dofmaster, using the D800 /700 table produces
400mm @ f2.8 @ 25m = 0.65m
135mm @ f2.8 @ 25m = 5.87m
(135mm @ f2 @ 25m = 4.12m)
(used 25m 'cos it's half a pitch width approx?)
Flickr
I could do this all in software and with math to be honest. One of the benefits of using shorter focal lengths is that it's a lot easier to get larger apertures. For instance, while I have a 70-200/F2.8 and there is a 200mm/F2 available, I own an 85mm F1.4 and could use that.
I won't have my assistant this week, and will be doing quite a lot of work on setup, so this may be an exercise best left for another time.
Thanks for the engaging conversation though.
perroneford@ptfphoto.com
I'd also taken on board the fact that faster lenses are available for the shorter focal lengths ( hence the 135 /f2 figure)
Another thought also crossed my mind ... whether AF performance would be affected in the scenarios being discussed ... ie long tele with small MP v shorter lens @ large MP for equivalent pixels on target.
By definition, there would be the same No of pixels available to the AF sensors, but that same area (No) of pixels would constitute a smaller region of the frame in overall terms, with the shorter lens.
I dont know enough about any of this to do anything other than wonder ... but in a world where subject isolation from bg is prized - especially pros like yourself ... I can't help but be puzzled:)
pp
btw
85mm @ 1.4 @ 25m = 7.48m
Flickr
And, the D200 sensor wasn't good.
An "accurate" reproduction of a scene and a good photograph are often two different things.
Perhaps not by today's standards, but I've gotten some nice shots with it.
perroneford@ptfphoto.com