I am using a 100-400mm 4.5-5.6 lens and a canon eos 50d. I am wondering if i switch to the sRAW (50% of pixels used) if I would get increased high iso noise reduction? At this point, I am using the camera with high iso noise reduction:strong and receiving less than adequate noise reduction. I have heard rumors that highlight tone priority will also cause high ISO noise. Is this true?
Your answers are appreciated
Posted by David Bostedo on February 1, 2010 at 11:53 PM
Erik - First, Highlight Tone Priority. Essentially what this does is underexpose your shot then apply tone curves to make up for the underexposure. Since you are increasing the level of an underexposed shot (the non-highlight parts anyway), this will generally lead to more noise in shadow areas. I'd prefer not to use HTP (I don't shoot Canon, so that's not an option) and instead just shoot to fit the histogram as far to the right as I can, then adjust during RAW processing.
As for sRAW, you are halving the overall resolution by only using half the pixels. That's all it does as far as I can tell. So you should get the exact same amount of noise. Shrinking an image (not just taking a smaller image to begin with) will always help noise since you're multiply sampling each pixel; But that's not what the camera is doing. If Canon were doing something clever like pixel binning (combining multiple sensor pixels into one image pixel) you could definitely get a noise advantage. However, I'm not sure that would be better than just shooting full RAW and then doing the down-sizing yourself. sRAW is basically just about saving space.
I'd say stick with full RAW, and expose to the right so that you're are just avoiding clipping the highlights, then you'll be capturing maximum info.
As for high ISO noise reduction, there's no free lunch. Any amount of noise reduction will lose you detail, but some are better than others. I don't know how good Canon's is in camera, but nothing is perfect. Noise is still changing the image at the sensor, and it's not possible to perfectly recreate the image no matter how good the noise reduction is.
Now when you say you are "receiving less than adequate noise reduction", what exactly are you expecting? There are different types of noise and approaches you can take. And if you're simply looking at 100% view on screen, you may want to consider if this accurately reflects what you want to print or show. Usually, 100% view can be misleading, since it doesn't match your final output.
That's a lot of stuff, so feel free to ask more questions. I'm sure I didn't adequately explain something in there. :)
Posted by Larry Fischer on February 2, 2010 at 12:33 PM
Get a Nikon D700. Compare below....
http://www.dxomark.com/
Posted by David Bostedo on February 2, 2010 at 12:58 PM
You DO have to be a little careful with DxO Mark and knowing what you're looking at. By default, they don't include resolution in the calculation. (The do if you look at the graphs and choose the "print" option.) And resolution helps alleviate noise by allowing for over-sampling at smaller reproduction sizes, and limiting the amount of up-sizing that has to happen from sensor image to final image.
That being said, the D3/D700 is going to be at least one stop better, and often closer to two, than the 50D at all ISOs, even accounting for the resolution difference. It's largely a property of the larger sensor collecting more light.
And currently, the D3s should have the best high ISO performance - almost a stop better than the D3/D700 from what I've seen. I don't know that it's on DxOMark yet though.
Posted by Larry Puckett on February 2, 2010 at 1:34 PM
You don't say what ISO you are using--this is a very important with the 50D as it does not do all that well with noise above ISO 400 IMHO especially in darker areas of an image. Also you might find that cutting your noise reduction back to normal or off and then using a post-processor noise reduction plugin like Noise Ninja would give you sharper images at reduced noise. Personally on my 50D I have noise reduction on low and turn off the HTP. I also keep long exposure noise reduction on Auto. This is all for landscapes so your needs may differ if as I assume you are after birds or other wildlife. One thing to keep in mind is that if you're trying to grab as many images of a BIF, then using in-camera noise reduction can slow down your processing speed and limit how many images you can record in a burst--Larry
Posted by David Bostedo on February 2, 2010 at 3:44 PM
I don't think high ISO NR applies to RAW files, either. So you're going to need to do noise reduction during processing if you're not working on JPGs.
Posted by Larry Puckett on February 2, 2010 at 5:36 PM
According to Rob Sheppard in his book on the 50D "Stronger noise reduction will increase the processing needed in camera (including with RAW)." I think the reason for this is that even though the NR is not applied to the RAW file the data are still stored with the file so it can be applied in DPP if the user chooses. There are a lot of settings in these cameras that get written with RAW files so they can optionally be applied in DPP.
Some folks like to shoot both RAW and JPEGS simultaneously and some attempting to capture action shots use the ability to use burst mode to get 6.3 FPS and up to 90 continuous shots with JPEGS and 10 shots with RAW. Stronger noise reduction will limit this due to the longer processing time. One thing that would be useful in all these manuals and books would be a section that just states what applies to JPEGS and what applies to RAW files--you usually have to really hunt for this info and in most it is not clear at all--Larry
Posted by Larry Fischer on February 3, 2010 at 7:54 AM
David,
I have a D300 and D700.
I have read where noise is high pixel sensors decreases as the size of the print decreases.
I have also experienced that there is no way to get a decent sized, low light photo from a d300 that matches the fewer pixels and low noise of the D700. I'm speaking of photos up to 24x18.
I have heard and seen on CSNP about the D3s.
the fisch
Posted by David Bostedo on February 3, 2010 at 9:15 AM
Larry - A couple of things...
-- Unless I'm missing something, the D300 and D3 are both 12 megapixel cameras, with the D300 having some small advantage (like 12.1 versus 12.3) that's really insignificant. You'd probably need to have about a 50% pixel increase for it to matter much in prints. (So a 15MP sensor the size of the D300's would noticeably help when compared to a D300.)
-- No matter how much you increased the D300's resolution, you probably wouldn't be able to match the D3 for noise (assuming the underlying circuitry and technology were similar). Overall sensor size will always help on the noise front because you are simply collecting more photons, so you can achieve a higher signal, and a better SNR, which leads to lower noise photos.
David, I think you are missing something. The d300(anything) is not a full size sensor like the D3 and D700. For a full size sensor and a and percentage sensor both having the same number of pixels, the full size sensor will win every time. There are two many posts and internet data explaining this for me to go into detail here. Check it out! No offense intended! It seems like you know the difference in your explanation which is kinda contrary to your opening remark! Maybe "I" am missing something!
Troup
Posted by David Bostedo on February 4, 2010 at 12:07 AM
Troup - No offense taken...I think you might be misunderstanding what I was saying. I'm very clear on the difference between full frame and smaller sensors, and the advantages of the full frame. What I was trying to clarify, and maybe did a bad job of, was three things :
1) Higher resolution can make up for higher pixel level noise. Assuming the same output size (final print or screen size) you will likely be downsizing the higher res image to match the lower res, which of course gives a sampling advantage mitigating potential pixel level noise due to higher resolution and smaller pixels. That's what the DxO mark article above deals with.
2) However, there's a limit here based on the overall signal collected for the whole sensor. A larger sensor simply collects more light (signal) and therefore - all else equal - will have lower noise than a smaller sensor for a given reproduction size.
So the "higher-res-mitigates-noise" effect only really applies to same-sized sensors - NOT when comparing a D300 to D700/D3. (I also should have said 18MP above instead of 15, since 18 would be 50% more than the D300, but that's a rough number anyway.)
3) All of this was in response to Larry's point that noise decreases with print size, and that he can't get as good a print from his D300 despite higher pixel count. I was saying the the D300 pixel count is not meaningfully higher, and that even if it were it would not be likely to match the overall noise capability of the full size sensor.
Hopefully, that's clearer.
Posted by Larry Fischer on February 4, 2010 at 11:36 AM
David and Troup,
Iron sharpens iron and one man sharpens another. I agree with David's last post.
Full sized sensors with the same number of pixels will do better than the same number of pixels in a smallers sensor. The larger pixel is a bigger bucket and the pixel receives more light than the smaller one--thus less noise.
I really like my D700 for low light. Hasselblad, as I understand it, has retained larger pixels even in their medium format. Not sure about other professional backs.
Are not pixels a comination of 4 "buckets"??? red, green and blue with one of them being duplicated??? green??
the fisch
Posted by David Bostedo on February 4, 2010 at 12:52 PM
Are not pixels a comination of 4 "buckets"??? red, green and blue with one of them being duplicated??? green??
Not really Larry. Each pixel in your camera DOES have a single color filter over it (unless you have a Foveon sensor). And there are twice as many green pixels as red or blue. (See Bayer Filter) But a single pixel on the sensor will also show up as a single pixel on the screen. The reason is that the final color of the screen pixel will be determined during de-mosaicing in the RAW converter based on it's own value and the pixels around it. This is true whether you do it yourself on the computer, or if it's done in camera to produce JPGs.
Now one misnomer this leads to is that sensors really only have 1/3 or 1/2 of the listed resolution. Some people think that a black-and-white sensor (which all of them are without the color filters and de-mosaicing) would thus have higher resolution. This really doesn't occur (black and white sensors have been made) and the de-mosaicing algorithms are so good now that you'd be hard pressed to say that anything is lacking with either the detail, or the color, resolution. That's why the Bayer filter array hasn't been supplanted by anything different yet - it's very successfully implemented.
Discussion
12 repliesYour answers are appreciated
As for sRAW, you are halving the overall resolution by only using half the pixels. That's all it does as far as I can tell. So you should get the exact same amount of noise. Shrinking an image (not just taking a smaller image to begin with) will always help noise since you're multiply sampling each pixel; But that's not what the camera is doing. If Canon were doing something clever like pixel binning (combining multiple sensor pixels into one image pixel) you could definitely get a noise advantage. However, I'm not sure that would be better than just shooting full RAW and then doing the down-sizing yourself. sRAW is basically just about saving space.
I'd say stick with full RAW, and expose to the right so that you're are just avoiding clipping the highlights, then you'll be capturing maximum info.
As for high ISO noise reduction, there's no free lunch. Any amount of noise reduction will lose you detail, but some are better than others. I don't know how good Canon's is in camera, but nothing is perfect. Noise is still changing the image at the sensor, and it's not possible to perfectly recreate the image no matter how good the noise reduction is.
Now when you say you are "receiving less than adequate noise reduction", what exactly are you expecting? There are different types of noise and approaches you can take. And if you're simply looking at 100% view on screen, you may want to consider if this accurately reflects what you want to print or show. Usually, 100% view can be misleading, since it doesn't match your final output.
That's a lot of stuff, so feel free to ask more questions. I'm sure I didn't adequately explain something in there. :)
http://www.dxomark.com/
DxO themselves have a good article on this.
That being said, the D3/D700 is going to be at least one stop better, and often closer to two, than the 50D at all ISOs, even accounting for the resolution difference. It's largely a property of the larger sensor collecting more light.
And currently, the D3s should have the best high ISO performance - almost a stop better than the D3/D700 from what I've seen. I don't know that it's on DxOMark yet though.
Some folks like to shoot both RAW and JPEGS simultaneously and some attempting to capture action shots use the ability to use burst mode to get 6.3 FPS and up to 90 continuous shots with JPEGS and 10 shots with RAW. Stronger noise reduction will limit this due to the longer processing time. One thing that would be useful in all these manuals and books would be a section that just states what applies to JPEGS and what applies to RAW files--you usually have to really hunt for this info and in most it is not clear at all--Larry
I have a D300 and D700.
I have read where noise is high pixel sensors decreases as the size of the print decreases.
I have also experienced that there is no way to get a decent sized, low light photo from a d300 that matches the fewer pixels and low noise of the D700. I'm speaking of photos up to 24x18.
I have heard and seen on CSNP about the D3s.
the fisch
-- Unless I'm missing something, the D300 and D3 are both 12 megapixel cameras, with the D300 having some small advantage (like 12.1 versus 12.3) that's really insignificant. You'd probably need to have about a 50% pixel increase for it to matter much in prints. (So a 15MP sensor the size of the D300's would noticeably help when compared to a D300.)
-- No matter how much you increased the D300's resolution, you probably wouldn't be able to match the D3 for noise (assuming the underlying circuitry and technology were similar). Overall sensor size will always help on the noise front because you are simply collecting more photons, so you can achieve a higher signal, and a better SNR, which leads to lower noise photos.
Troup
1) Higher resolution can make up for higher pixel level noise. Assuming the same output size (final print or screen size) you will likely be downsizing the higher res image to match the lower res, which of course gives a sampling advantage mitigating potential pixel level noise due to higher resolution and smaller pixels. That's what the DxO mark article above deals with.
2) However, there's a limit here based on the overall signal collected for the whole sensor. A larger sensor simply collects more light (signal) and therefore - all else equal - will have lower noise than a smaller sensor for a given reproduction size.
So the "higher-res-mitigates-noise" effect only really applies to same-sized sensors - NOT when comparing a D300 to D700/D3. (I also should have said 18MP above instead of 15, since 18 would be 50% more than the D300, but that's a rough number anyway.)
3) All of this was in response to Larry's point that noise decreases with print size, and that he can't get as good a print from his D300 despite higher pixel count. I was saying the the D300 pixel count is not meaningfully higher, and that even if it were it would not be likely to match the overall noise capability of the full size sensor.
Hopefully, that's clearer.
Iron sharpens iron and one man sharpens another. I agree with David's last post.
Full sized sensors with the same number of pixels will do better than the same number of pixels in a smallers sensor. The larger pixel is a bigger bucket and the pixel receives more light than the smaller one--thus less noise.
I really like my D700 for low light. Hasselblad, as I understand it, has retained larger pixels even in their medium format. Not sure about other professional backs.
Are not pixels a comination of 4 "buckets"??? red, green and blue with one of them being duplicated??? green??
the fisch
Not really Larry. Each pixel in your camera DOES have a single color filter over it (unless you have a Foveon sensor). And there are twice as many green pixels as red or blue. (See Bayer Filter) But a single pixel on the sensor will also show up as a single pixel on the screen. The reason is that the final color of the screen pixel will be determined during de-mosaicing in the RAW converter based on it's own value and the pixels around it. This is true whether you do it yourself on the computer, or if it's done in camera to produce JPGs.
Now one misnomer this leads to is that sensors really only have 1/3 or 1/2 of the listed resolution. Some people think that a black-and-white sensor (which all of them are without the color filters and de-mosaicing) would thus have higher resolution. This really doesn't occur (black and white sensors have been made) and the de-mosaicing algorithms are so good now that you'd be hard pressed to say that anything is lacking with either the detail, or the color, resolution. That's why the Bayer filter array hasn't been supplanted by anything different yet - it's very successfully implemented.