Do you guys dumb down your RAW pixels?

JayClark79

The Pro Photographer that im interning with doesnt shoot his 5d mark ii at full res Raw files... he shoots them at Raw 2 or something like that he said it makes it something like 10 megapixels... he says he does this to save on memory card space and that the 21 mega pixels it typically has isnt necessary....

Do you do this? :scratch

ziggy53

If you aren't planning to print larger than an 8" x 10", or so, the 5D MKII at sRAW1 is plenty. For many subjects that image size is enough for a 20" x 30". For candids even the sRAW2 setting works fine.

It all depends upon your intended application of the file.

For any sort of formals and portraiture I would always use the largest files size and largest resolution. Same for fine art images. For many other subjects and uses, it would depend. I appreciate the options.

It partly too depends upon the RAW processor you use. Some RAW image processing software takes much longer to process the smaller RAW files than the full sized files, odd as that seems.

Art Scott

no to the dumbing down of raw files........the only dumbing down is by cropping into the aspect ratio
I normally use and that is 4x5 at somewhere between 450 and 300 dpi.......I lose enuff when cropping
that I do not need to loose any pixels by shooting less to start with.........

Overfocused

I'd rather use full size jpeg than half sized RAW, large format prints are much more common nowadays

ThatCanonGuy

I'll second Richy. You never know how big you might want to print an image in the future. I mean, seriously, to save memory sounds like the least important reason to shoot at less then full res. If somebody says that, I tell them to go to Costco or wherever and buy 32gigs for 100 bucks.

Moogle Pepper

For me, I prefer keeping full raw, for weddings, on prep to first dance shots. Once that is over and it is just reception, sraw1. Personal work it is mainly full raw, travel sraw1.

And yes I printed huge on 8 mp images, 14 x 24 (odd ratio, I know.) and 16 x 24s from sraw1 sized images.

Art Scott

richy wrote: »

I remember my bugbear with 8mp cameras and 35mm film was it just didn't enlarge past 11x14 with any real quality. I know people say the bigger the print the more the viewing distance but in my experience b&g will look closely at the print.
Not trying to wail on anyone, just curious

This is where Genuine Fractals and I have become best buds......when shooting with my 6mp KM 7D........full raw, processed in LR and PSCS and then uprezed to 30x40 at 300dpi......this gave me a huge print with absolutely no viewing distance needed and to this day I still decided what the largest print I want to sell and uprez and then do final sharpening with unsharp mask and then save........I do have some huge jpgs but storage is cheap........

Matthew Saville

Very, very happy with my 12 megapixel Nikons. NOT looking forward to the next generations and their inevitable megapixel boosts.

I make massive prints all the time with 12 megapixels, they look just fine.

I shoot about 100,000 images per year and keep 20-40K of them. I do the math on Nikon's RAW file sizes, and I just don't have the extra time or cash to support the difference. Or if I do, I'd rather spend the money on other things.

I'll probably be using the D700 / D300 / D3s as long as I possibly can. Can't wait for the D4 to come out, then eventually the D3s will be $2500 or so... :-D

=Matt=

Moogle Pepper

richy wrote: »

Do you find it an issue with enlargements? (I promise its not a loaded question, genuinely interested in how that looks on something in the 30x40 range).

Also have you tried shooting 'full'raw at any point? Given how cheap computers and storage and cards are I'm just curious how much of an advantage it is to shoot sraw. If I sell three canvasses or large prints a year it covers my budget for computers \ storage and card replacement for the year.

I remember my bugbear with 8mp cameras and 35mm film was it just didn't enlarge past 11x14 with any real quality. I know people say the bigger the print the more the viewing distance but in my experience b&g will look closely at the print. Thats why I shot MF film alongside digital and then moved to the 5d2. If I only needed 10mp I would get a brace of 1d3's rather than downsize raws. For me the point of the 5d2 is the IQ and image size, throwing away half of that plus would just suggest I needed a different camera? It screams 1d3 or d700, smaller files and better AF thrown in.

Not trying to wail on anyone, just curious

The biggest I have printed with files from an 8 mp image is16 x 24, which looks really good. I'll let you know how a 24 x 36 looks when I send one out to the lab. I don't have the money to always make huge prints.

I do shoot full raw, but at only key times, as I have mentioned. I can live with just 12 mp, or whatever sRAW1 drops the MP to.

billythek

I wouldn't argue with a guy who actually does this for a living. If it works for him, more power to him.

For my purposes, I just shoot everything maximum res in RAW, import straight to Lightroom, and cull extensively there. But I don't have to deal with the volume of photos that a pro wedding shooter does. He has to keep, and archive a lot more of his shots than I do.

Matthew Saville

richy wrote: »

I shoot around the same as Matt give or take and keep them all. The maths for the 5d2 assuming 26 ish average file sizes for 100k shots a year is 2.6TB of files, keeping one copy on external drives would cost approx $140. A dvd copy would run about $120. I do 2x external drives and one to dvd so my yearly storage bill is around $400.
Other pros make other choices and obviously aren't crazy. Just work differently. A fast pc every other year is about $1000 - $1200.
Its really hard to quantify time involved. Archiving to external drives is quick. I tend to archive to dvd when I'm working the next weddings files so not really much time there. For time actually working the files, it probably does take somewhere like 30% longer but I shot 3 weddings on 10-10-10 and was all done editing my lunch the next day so its not a huge drag time wise. Others may shoot more shots per hour?

The comments re fractals are very interesting, perhaps in another thread we can see how good it is with a test? We pick a real world raw file and compare fractals against say bicubic / printer dithering and maybe some PS wizzardry? dont want to hijack this thread anymore re that

Especially as an Apple user, it all adds up. $100 here, $100 there, $1000 here, $2000 there... I would definitely be switching back to PC if I were to try and embrace a doubly large filesize, that's for sure. (Or building an E-FIX Mac, more likely. Just can't stand Windows...)

Maybe some day when I'm doing very well I'll be able to just go out and buy the latest high-res cameras, and new RAM / CF / HDD etc. every year or two. But until then, the only point I want to make is that I'm not missing out. Okay I might be missing out on not being rich, but that has nothing to do with my megapixels ... ;-)

billythek wrote: »

I wouldn't argue with a guy who actually does this for a living. If it works for him, more power to him.

For my purposes, I just shoot everything maximum res in RAW, import straight to Lightroom, and cull extensively there. But I don't have to deal with the volume of photos that a pro wedding shooter does. He has to keep, and archive a lot more of his shots than I do.

As a hobbyist landscape photographer, I can absolutely understand and relate to this mentality. If I were a professional landscape, nature, or wildlife photographer, I'd be all over EVERY new increase in megapixels, in fact I might have switched from a D300 to the 7D by now, who knows. Canon has still got the ONLY 70-200 f/4's on the market, and that would be an absolute must in my bag as a wildlife photographer.

But my point is, 75% of the time with the images I'm shooting, resolution just doesn't matter beyond ~10 megapixels. So even if I could afford the extra $2000-3000 per year that it'd require to jump to a 24 megapixel Nikon or something, I'd rather save the cash or at least spend it on the people I love, etc...

=Matt=

studio1972

I shoot full raw for everything, but have considered the SRAW as I'm sure that it would be fine. Photo storage is less of a problem than video anyway however. In the long term (several months post wedding) the files get wiped from my computer and stored as JPEGs on DVD and online, I don't see any need to keep the RAW files once all the processing is done and the client has accepted them.

Matthew Saville

studio1972 wrote: »

I shoot full raw for everything, but have considered the SRAW as I'm sure that it would be fine. Photo storage is less of a problem than video anyway however. In the long term (several months post wedding) the files get wiped from my computer and stored as JPEGs on DVD and online, I don't see any need to keep the RAW files once all the processing is done and the client has accepted them.

Just from one wedding pro to another, I'd strongly suggest at least keeping the RAW files to your topmost portfolio pieces. Just trust me on this one! I cannot tell you how many times I've gone back to files from weddings 3-5 years ago and thought "OMG, I processed it like THAT?"

=Matt=

JayClark79

thanks guys... iv seen a pig 30+ inch canvas of his that he shot at raw1 in full daylight and it looks great..... however when i look at some of his low light images i feel they have more noise then mine and im currently only shooting a T1i which is 15 mega pixels... i asked him about it and i dont think he half believed me.... but if you ask me dumbing down the pixels... + low light... would give higher noise right?

ziggy53

JayClark79 wrote: »

thanks guys... iv seen a pig 30+ inch canvas of his that he shot at raw1 in full daylight and it looks great..... however when i look at some of his low light images i feel they have more noise then mine and im currently only shooting a T1i which is 15 mega pixels... i asked him about it and i dont think he half believed me.... but if you ask me dumbing down the pixels... + low light... would give higher noise right?

A Canon 5D MKII should have demonstrably better random noise at the same ISO setting compared to a Canon T1i. The visible noise should be even a little better using sRAW1 because it does pixel binning to reduce the image size to 10 MPix.

Matthew Saville

ziggy53 wrote: »

A Canon 5D MKII should have demonstrably better random noise at the same ISO setting compared to a Canon T1i. The visible noise should be even a little better using sRAW1 because it does pixel binning to reduce the image size to 10 MPix.

Pixel binning, naughty word!!! :-P

ziggy53

richy wrote: »

Do canon have true pixel binning for stills? Thats awesome news. It might make using sraw for reception shots more attractive!

...

While it appears that the Canon 5D MKII does use a binning approach to sRAW files, as opposed to discarding pixels, the results are not quite as good as shooting full sized RAW and then using a decent RAW converter to "normalize" the file down to the lesser pixels.

There is a pretty good comparison of RAW/downsized and compared to sRAW about 2/3rds of the page down here:

http://rolandlim.wordpress.com/2008/12/23/canon-eos-5d-mark-ii-review/

Note that in the above review he resamples the hires file in PS, rather than in the RAW converter. He finds the results of RAW/downsized to sRAW sizes, compared to shooting sRAW, too close to call.

NeilL

richy wrote: »

Reducing pixels if done by binning should reduce noise as the same size in relation to the sensitivity and noise of the a2d process is increased. i.e. if you have a margin of error of +or- 1 in a sample of 10 versus a sample of 20, the latter would be twice as accurate ergo less deviation. Noise is also relative, smaller pixels equals more noise but that noise has less effect as it is diluted more in the output as the resolution is higher (if that makes any sense lol). Either way, its all good

Please, Richy, if you're gonna write something marvelously insightful like this could you also write it in English!rolleyes:D

As far as I understand things, noise is fixed collateral damage of whatever system is processing data. Noise is a permanent imprint on an image. It is not negotiable after the fact. It can only be masked. Noise can be more or less detracting, more or less easily tamed, but it cannot be rinsed away by changing the image statistics. Magic is not yet offered along with smile recognition and video.

Neil

studio1972

NeilL wrote: »

Please, Richy, if you're gonna write something marvelously insightful like this could you also write it in English!rolleyes:D

As far as I understand things, noise is fixed collateral damage of whatever system is processing data. Noise is a permanent imprint on an image. It is not negotiable after the fact. It can only be masked. Noise can be more or less detracting, more or less easily tamed, but it cannot be rinsed away by changing the image statistics. Magic is not yet offered along with smile recognition and video.

Neil

What Richy said was actually correct. As an image is down scaled (reduced in resolution) the noise reduces. That's why often high ISO images look great on the web at 1/100 of their original resolution, but not so great when you zoom into the original at 100%.

studio1972

richy wrote: »

Physics rarely translates into English and canon have been putting pixie dust on their sensors for years, how else could their 18mp apsc sensor have so little noise compared to a Nikon (operation cat amongst pidgeons complete, time to hide).
But seriously, you are partially right. Once an image is completed the noise is there. What pixel binning does is step in before that point. Each pixel has a lens that focuses light into a well. That well builds up a charge based on the amount of light hitting it. This analogue reading is then digitised and the efficency of this bit (plus the sensitivity of the sensor and the size of the pixel itself) determines for the most part the amount of noise. Pixel binning (i believe) combines the readings from wells prior to conversion rather then after when 'noise' becomes cemented in the file.

So you shoot a uniform red picture and the red channel for sake or argument reads 100 abritrary units plus or minus 1. So a 2 percent deviation / noise across the board. Now say you employ 2 to 1 binning, you now have a reading of 200 plus or minus 1, so 1 percent noise. You just sacrifice resolution. It does work, and you aren't wrong but it works earlier in the sequence so what you said and it working aren't mutually exclusive.

It doesn't actually make any difference whether you combine the pixels digitally or through some analogue process as you suggest (although I'm sure it would be digital in the case of a camera sensor). As Ziggy already mentioned it is possible to do a better job of this in software than the in-camera system (but the in-camera system gives you the benefit of more images per card etc.).

ziggy53

It is not known exactly how Canon creates the sRAW files, but the "binning" effect of resampling is apparent in the results. The other way to reduce pixel count is to discard pixels, and some video cameras appear to do that as evidenced by their failure to reduce sensor noise. (Discarding pixels keeps the same noise signature as is in the full resolution original.)

The problem with binning techniques and technologies are multifold. Since Canon does not describe their process, we can only guess as to their methods. Since we don't have any control over the process I submit that only the results are pertinent to a worthwhile conversation.

The results of Canon's sRAW are actually very good and very close to what you get with the best high-powered computer and software. That's very good news. I do believe that a full-blown modern computer and software can do better, but that's mostly related to the more accurate number systems that a full computer and software allow.

What it means is if the output resolution of the sRAW is sufficient for your needs, the full control of RAW processing is available as well. As compared to a medium JPG this is excellent as it allows considerable post-processing and very good detail along with reduced intrinsic sensor noise and smaller file sizes.

For an 8" x 10" that did not need cropping from the original and if the subject matter does not contain too much detail, the sRAW2 files, with 5 MPix resolution, is often enough. The sRAW1 is even more capable with enough native resolution for a true 1:1 pixel to pixel relationship in 300 dpi prints. (3000 x 2400 pixels, or 7.2MPixels is what's required for a 1:1 relationship. A Canon 5D MKII sRAW1 file has 3861 x 2574 pixels, so more than enough for the 8x10 crop.)

As far as vertical banding and aliasing (jaggies) those issues are largely addressed in early firmware and software updates. Any remaining issues seem to be related to sharpening technique and particular color boundaries.

NeilL

richy wrote: »

Physics rarely translates into English

Wouldn't agree. Rather, I'd say that no physics can be known unless it can be expressed in language (math also being a language, and you need English, or some language, to do math too).

I wonder if your use of the terms error and noise as synonymous is correct. Error would seem to me to mean something like the non-preservation of accuracy in the process of photon reading through digital conversion and beyond. For example, each individual pixel's performance would be variable for every instance of capture, and that variability is the measure of error. Noise is data added to the signal, synthetic data produced by the system's physical components, which to some extent confuses the signal. Error is the variability of signal data, which is the variability of performance of the system components in reading and representing light data. Noise is the signal intrinsic to the system which adds to the signal captured. They are two different things, I think. Their remedy is also somewhat different. Binning is a technique which involves appropriate clocking, voltage and charge capacity technology. It amplifies certain aspects of the signal such as exposure and dynamic range, and thus increases the signal to noise ratio, but it does not lower noise. It also diminishes certain signal aspects such as resolution. Some noise, such as dark current noise, will always cause an intractable limit on the increase of the signal to noise ratio by binning.

So binning and its mathematical description can seem glibly magical, but in fact it is a compromise in the push and pull between ideals and materials.

Neil

ziggy53

Pixel "binning" is simply the colloquial name given to interpolation used for down-rezzing images, sometimes also a part of "normalization" of the image to a smaller dimensional size. (Normalization of all of the data is not what is being discussed, but normalization of data subsets is a part of the process.)

Since more pixels are sampled in the interpolation process for the input than output, some reduction of random sensor noise results from the data averaging that occurs as a result of the sampling algorithms "and" the reduction in gross pixel count. It is the same exact thing as using competent software interpolation to reduce image size on a computer, with the exception that a computer probably has more memory in which to perform more accurate interpolation, as well as more time and processor power in which to sample the data with higher precision.

This is not the same as "noise reduction" even though reduced noise is a predictable by-product of the process.

NeilL

ziggy53 wrote: »

Pixel "binning" is simply the colloquial name given to interpolation used for down-rezzing images, sometimes also a part of "normalization" of the image to a smaller dimensional size. (Normalization of all of the data is not what is being discussed, but normalization of data subsets is a part of the process.)

Since more pixels are sampled in the interpolation process for the input than output, some reduction of random sensor noise results from the data averaging that occurs as a result of the sampling algorithms "and" the reduction in gross pixel count. It is the same exact thing as using competent software interpolation to reduce image size on a computer, with the exception that a computer probably has more memory in which to perform more accurate interpolation, as well as more time and processor power in which to sample the data with higher precision.

This is not the same as "noise reduction" even though reduced noise is a predictable by-product of the process.

Yep, the discussion has been a bit imprecise, some people talking about digital capture hardware, error and noise, signal to noise ratio, others talking about downresing with software in camera or out.

Anyway, in the case of the latter the net result of averaging you describe is the increase of signal to noise ratio, not the disappearance of noise, in that noise is more likely to be constant compared with signal, therefore the summing of noise can be avoided, while the summing of signal data strengthens the ratio of signal over noise.

Neil

NeilL

richy wrote: »

Isn't semantics fun.

Do you mean you can understand stuff without it?

Neil

NeilL

richy wrote: »

I barely understand it with it, without it is much more fun. I used to work for a large petrochem company in R&D (designing catalysts but mostly seeing how long I could keep my finger in liquid nitrogen) but that was a long time ago. I make a passing attempt to grasp the overall concepts for as much as I need to for my current job as a photographer but I am happy to learn more, wine and old age have obviously destroyed the majority of the grey matter.

I had 'dumbed down' the language at your request but obviously that wasn't needed which confuses me a little as to your earlier statement. You obviously understand the situation and have stated it quite elegantly so thank you

Maybe Richy it was your swashbuckling style that was distracting me, making me have pictures of swordfights on the quarter deck with pegleg pirates, rather than electrons being shunted from pixel wells through parallel and series shift registers!:D

Neil

Matthew Saville

In an effort to get back on topic, I will say:

I'm going through a 2nd shooter's images from the 7D, and I believe they shot in sRAW1, and WOW do the images have horizontal banding issues. Around edges of color, it is extremely visible. Caused by pixel binning? I dunno, but I'm having second thoughts about dumbing down my RAW pixels. Maybe there's a REASON Nikon hasn't yet come out with smaller RAW size even in their 24 megapixel and 16 megapixel cameras...

=Matt=

Nikolai

With HDD space going for about $50..$80 per 1Tb I really could care less if I shoot 1,500 frames per day for a public event, or a couple hundres frames for a studio shoot, or just a dozen or two when hunting the Green Flash.
Why worry over such a small matter and risk a potentially great capture being taken in low-res?

Matthew Saville

Nikolai wrote: »

With HDD space going for about $50..$80 per 1Tb I really could care less if I shoot 1,500 frames per day for a public event, or a couple hundres frames for a studio shoot, or just a dozen or two when hunting the Green Flash.
Why worry over such a small matter and risk a potentially great capture being taken in low-res?

...As we discussed earlier, hard drives are only a small part of the equation. At the end of the day, if the average photographer were to shoot every image un-compressed RAW, 14-bit, 20+ megapixels, ...it would cost hundreds or thousands of dollars to maintain the same speed of workflow compared to say for example a 12 megapixel JPG shooter.

By the logic of "bigger is better", everybody ought to be shooting Nikon's in-camera TIF files; they're 40 MB for a 12 megapixel image! Um, no thank you...

The best thing a photographer can do is to understand ALL image capture modes, and use the right settings for the right applications. Bottom line- I wouldn't want to be caught WITHOUT the ability to shoot one or the other, depending on the situation...

=Matt=

Nikolai

Matthew Saville wrote: »

...As we discussed earlier, hard drives are only a small part of the equation. At the end of the day, if the average photographer were to shoot every image un-compressed RAW, 14-bit, 20+ megapixels, ...it would cost hundreds or thousands of dollars to maintain the same speed of workflow compared to say for example a 12 megapixel JPG shooter.

By the logic of "bigger is better", everybody ought to be shooting Nikon's in-camera TIF files; they're 40 MB for a 12 megapixel image! Um, no thank you...

The best thing a photographer can do is to understand ALL image capture modes, and use the right settings for the right applications. Bottom line- I wouldn't want to be caught WITHOUT the ability to shoot one or the other, depending on the situation...

=Matt=

Matt,
I shoot full RAW (21mp) plus small JPEG. That combo plus the right equipment and the right tools gets me a fairly decent speed workflow-wise (nobody ever beat me to an upload ).
And I really have no clue where do "hundreds or thousands of dollars" come from - I think I would notice.
I may buy a new HDD every now and then (like once in 18 months), but that's nothing compared to the price of a body upgrade or a decent glass.

As I said, small matters...

Shooting TIFFs - well, I don't understand it, but if it works for somebody - by all means.

ziggy53

Matthew Saville wrote: »

In an effort to get back on topic, I will say:

I'm going through a 2nd shooter's images from the 7D, and I believe they shot in sRAW1, and WOW do the images have horizontal banding issues. Around edges of color, it is extremely visible. Caused by pixel binning? I dunno, but I'm having second thoughts about dumbing down my RAW pixels. Maybe there's a REASON Nikon hasn't yet come out with smaller RAW size even in their 24 megapixel and 16 megapixel cameras...

=Matt=

Be sure to try another RAW converter to see if the issue is occurring in the converter.

Also make sure that the shooter is using the latest firmware. There was a vertical magenta banding issue that was corrected in earlier firmware. I'm not seeing anyone else talking about "horizontal" banding with 7D sRAW files however.