how to see if CR2 Long Exposure Noise Reduction has been applied?

[niels123]

Just by looking at a CR2 file or its metadata, is it possible to see if the in-camera long exposure noise reduction was set to 'on' or 'off'?

 

[tron]

Just by looking at a CR2 file or its metadata, is it possible to see if the in-camera long exposure noise reduction was set to 'on' or 'off'?

By using DPP you can know just the setting of your camera which is usefull only when it is off or on. If it is auto (the default I believe) then I do not know.

 

[niels123]

Just by looking at a CR2 file or its metadata, is it possible to see if the in-camera long exposure noise reduction was set to 'on' or 'off'?

By using DPP you can know just the setting of your camera which is usefull only when it is off or on. If it is auto (the default I believe) then I do not know.

Do you know what the default factory setting is when you buy your 5D III new?

What does auto mean? If it is set to auto, when does / doesn't the camera apply the LENR?

 

[neuroanatomist]

The setting is recorded in the metadata, but not as a unique item (it's coded in the custom controls field as a 0/1/2 somewhere in that string of numbers, the position depends on the camera model). As tron states, the setting (Auto, Enable, Disable) will be shown in DPP but that won't tell you whether or not NR was applied if Auto was set. FYI, the default setting is Disable.

Note that when enabled, it only applies for 1 s or longer exposures (and in auto, it's applied if the camera thinks there's noise to reduce). Also, Canon notes that at ISO 1600 and higher, LENR may add noise (which is true - with random noise, subtracting is the same as adding). What LENR does is remove hot/stuck pixels, and that's something that most RAW converters do automatically. So, my recommendation is to just leave it off.

 

[tron]

Just by looking at a CR2 file or its metadata, is it possible to see if the in-camera long exposure noise reduction was set to 'on' or 'off'?

By using DPP you can know just the setting of your camera which is usefull only when it is off or on. If it is auto (the default I believe) then I do not know.

Do you know what the default factory setting is when you buy your 5D III new?

What does auto mean? If it is set to auto, when does / doesn't the camera apply the LENR?

I had read somewhere (but I do not remember where) if the speed is longer than 1 Sec but that is not enough. It is also a function of ISO which will have to be really high. Also, it can also be a function of available light. I am sorry but I do not know the algorithm.

 

[niels123]

The setting is recorded in the metadata, but not as a unique item (it's coded in the custom controls field as a 0/1/2 somewhere in that string of numbers, the position depends on the camera model). As tron states, the setting (Auto, Enable, Disable) will be shown in DPP but that won't tell you whether or not NR was applied if Note that when enabled, it only applies for 1 s or longer exposures (and in auto, it's applied if the camera thinks there's noise to reduce). Also, Canon notes that at ISO 1600 and higher, LENR may add noise (which is true - with random noise, subtracting is the same as adding). What LENR does is remove hot/stuck pixels, and that's Auto was set. FYI, the default setting is Disable.

something that most RAW converters do automatically. So, my recommendation is to just leave it off.

If you leave it to off, take a shot at ISO 1600 and 30s exposure with the LENR off and then manually substract a dark frame with pixel fixer, is there then a 'mathematical' way to compare the amount of noise between the origional and substracted frame other than by visually comparing? In reviews, graphs with quantize the amount of noise at various iso settings are often shown so it should be possible, right?

 

[neuroanatomist]

If you leave it to off, take a shot at ISO 1600 and 30s exposure with the LENR off and then manually substract a dark frame with pixel fixer, is there then a 'mathematical' way to compare the amount of noise between the origional and substracted frame other than by visually comparing? In reviews, graphs with quantize the amount of noise at various iso settings are often shown so it should be possible, right?

Sure, it's possible to quantify noise. Backing up a step, what are you trying to accomplish?

Subtracting a dark frame affects only fixed noise (hot pixels, pattern noise). As I stated, subtracting a dark frame actually adds random noise (it does so at all ISO values, but at lower ISO the noise is so minimal that adding 41% more of it isn't noticed). The way to reduce random noise is to average images.

 

[niels123]

If you leave it to off, take a shot at ISO 1600 and 30s exposure with the LENR off and then manually substract a dark frame with pixel fixer, is there then a 'mathematical' way to compare the amount of noise between the origional and substracted frame other than by visually comparing? In reviews, graphs with quantize the amount of noise at various iso settings are often shown so it should be possible, right?

Sure, it's possible to quantify noise. Backing up a step, what are you trying to accomplish?

Subtracting a dark frame affects only fixed noise (hot pixels, pattern noise). As I stated, subtracting a dark frame actually adds random noise (it does so at all ISO values, but at lower ISO the noise is so minimal that adding 41% more of it isn't noticed). The way to reduce random noise is to average images.

There's two types of photography in which I use long exposure and don't want to wait for a dark frame after each exposure:

1) High-magnification indoor macro photography with high CR daylight bulbs. Generally, I have 1 - 10 seconds of exposure per frame, always at 100 iso
2) Sometimes I like to 'play' a bit with night landscapes that include the heaven. In this case I just take 30s exposure frames without LENR, because I can do that at home. The iso varies between 400 and 1600 and if it's possible to quantify noise, than it's also possible to see if it makes sense to substract a dark frame or not. It may go a bit extreme, but you will get the best result if you try both and compare for each individual case.

 

[tron]

If you leave it to off, take a shot at ISO 1600 and 30s exposure with the LENR off and then manually substract a dark frame with pixel fixer, is there then a 'mathematical' way to compare the amount of noise between the origional and substracted frame other than by visually comparing? In reviews, graphs with quantize the amount of noise at various iso settings are often shown so it should be possible, right?

Sure, it's possible to quantify noise. Backing up a step, what are you trying to accomplish?

Subtracting a dark frame affects only fixed noise (hot pixels, pattern noise). As I stated, subtracting a dark frame actually adds random noise (it does so at all ISO values, but at lower ISO the noise is so minimal that adding 41% more of it isn't noticed). The way to reduce random noise is to average images.

There's two types of photography in which I use long exposure and don't want to wait for a dark frame after each exposure:

1) High-magnification indoor macro photography with high CR daylight bulbs. Generally, I have 1 - 10 seconds of exposure per frame, always at 100 iso
2) Sometimes I like to 'play' a bit with night landscapes that include the heaven. In this case I just take 30s exposure frames without LENR, because I can do that at home. The iso varies between 400 and 1600 and if it's possible to quantify noise, than it's also possible to see if it makes sense to substract a dark frame or not. It may go a bit extreme, but you will get the best result if you try both and compare for each individual case.

This thread is getting very interesting. I shoot landscape astrophotography and my most common setting is 20-25 sec f/2.8 at ISO 10000. I set LENR to ON but with these settings at night AUTO=ON practically. I have turned it to OFF a few times to shoot timelapses. The individual photos didn't seem different than the ones where it is set to ON but then there were so many stars visible that it wasn't easy to spot a few hot pixels. The read noise reference is interesting.

 

[neuroanatomist]

Got it. So, you want to empirically test in addition to 'trusting your eyes'. In both situations you describe, I'd leave LENR off.

I've used aMatLab script to measure image noise, but that's not something most people have available. Crudely, you can just estimate noise from the histogram, e.g. http://www.cambridgeincolour.com/tutorials/image-noise-2.htm.

I shoot landscape astrophotography and my most common setting is 20-25 sec f/2.8 at ISO 10000. I set LENR to ON but with these settings at night AUTO=ON practically. I have turned it to OFF a few times to shoot timelapses. The individual photos didn't seem different than the ones where it is set to ON but then there were so many stars visible that it wasn't easy to spot a few hot pixels. The read noise reference is interesting.

FYI, the 41% increase from subtracting a single dark frame is just math - noise increases as √2 = 1.41. Try shooting a manual dark frame (same exposure/ISO but with the lens cap on), no LENR. Open it in your favorite RAW converter, convert to JPG and check for hot pixels (whether you see them on the RAW converter depends on the converter - for example, DxO only removes them when you view at 100%, but they're gone in the JPG automatically).

 

[niels123]

Got it. So, you want to empirically test in addition to 'trusting your eyes'. In both situations you describe, I'd leave LENR off.

I've used aMatLab script to measure image noise, but that's not something most people have available. Crudely, you can just estimate noise from the histogram, e.g. http://www.cambridgeincolour.com/tutorials/image-noise-2.htm.

I shoot landscape astrophotography and my most common setting is 20-25 sec f/2.8 at ISO 10000. I set LENR to ON but with these settings at night AUTO=ON practically. I have turned it to OFF a few times to shoot timelapses. The individual photos didn't seem different than the ones where it is set to ON but then there were so many stars visible that it wasn't easy to spot a few hot pixels. The read noise reference is interesting.

FYI, the 41% increase from subtracting a single dark frame is just math - noise increases as √2 = 1.41. Try shooting a manual dark frame (same exposure/ISO but with the lens cap on), no LENR. Open it in your favorite RAW converter, convert to JPG and check for hot pixels (whether you see them on the RAW converter depends on the converter - for example, DxO only removes them when you view at 100%, but they're gone in the JPG automatically).

Thanx
I haven't used mathlab for years (I studied chemistry and abandoned it as quickly as possible ), but I'm curious in checking the script and see if I can get it to work here. Could I get it from you?

Where in the exif of a CR2 file do I find the 0/1/2 setting for LENR?

 

[neuroanatomist]

I haven't used mathlab for years (I studied chemistry and abandoned it as quickly as possible ), but I'm curious in checking the script and see if I can get it to work here. Could I get it from you?

One matlab example: http://www.mathworks.com/matlabcentral/fileexchange/45940-noise-level-estimation

Where in the exif of a CR2 file do I find the 0/1/2 setting for LENR?

There's not a separate field for it that I can see in exiftool, but it must be in the EXIF because DPP shows the setting. I believe (but I'm not positive) that it's specified in the Custom Controls field. I suspect the order of those is specific to the camera model, since different customization options are available on different models. You could figure it out by taking three pictures in series changing only that setting to each of the three options, and see which number changes.

From my 1D X, the field looks like this:

Custom Controls: 0 1 0 0 0 0 0 0 0 0 0 7 0 8 0 1 0 1 0 1 0 3 0 0 0 0 0 0 0 0 255 255 0 3 112 48 0 0 0 0 0 0 0 0 0 0 0 20 80 0 0 0 7 0

 

[niels123]

I haven't used mathlab for years (I studied chemistry and abandoned it as quickly as possible ), but I'm curious in checking the script and see if I can get it to work here. Could I get it from you?

One matlab example: http://www.mathworks.com/matlabcentral/fileexchange/45940-noise-level-estimation

Where in the exif of a CR2 file do I find the 0/1/2 setting for LENR?

There's not a separate field for it that I can see in exiftool, but it must be in the EXIF because DPP shows the setting. I believe (but I'm not positive) that it's specified in the Custom Controls field. I suspect the order of those is specific to the camera model, since different customization options are available on different models. You could figure it out by taking three pictures in series changing only that setting to each of the three options, and see which number changes.

From my 1D X, the field looks like this:

Custom Controls: 0 1 0 0 0 0 0 0 0 0 0 7 0 8 0 1 0 1 0 1 0 3 0 0 0 0 0 0 0 0 255 255 0 3 112 48 0 0 0 0 0 0 0 0 0 0 0 20 80 0 0 0 7 0

Just wondering: the only thing you're interested in substracting is the sensor noise, not the random noise that is also there. So: would it be possible to first average out a series of dark frames and then substract the averaged dark frame from your photo?

 

[neuroanatomist]

So: would it be possible to first average out a series of dark frames and then substract the averaged dark frame from your photo?

That's a very commonly used technique in astrophotography.

 

[niels123]

So: would it be possible to first average out a series of dark frames and then substract the averaged dark frame from your photo?

That's a very commonly used technique in astrophotography.

For *good* results, how many dark frames would you average out for 100, 200, 400, 800 and 1600 iso?

Are the dark frames specific to an individual camera or can they be used for all cameras that are the same type (e.g. 5D III).

 

[Valvebounce]

Hi Neils.
My understanding is that the dark frames are only specific to the temperature of the camera at the time of the shoot, I think that the noise is relative to the sensor temperature.

Cheers, Graham.

For *good* results, how many dark frames would you average out for 100, 200, 400, 800 and 1600 iso?

Are the dark frames specific to an individual camera or can they be used for all cameras that are the same type (e.g. 5D III).

 

[tron]

Got it. So, you want to empirically test in addition to 'trusting your eyes'. In both situations you describe, I'd leave LENR off.

I've used aMatLab script to measure image noise, but that's not something most people have available. Crudely, you can just estimate noise from the histogram, e.g. http://www.cambridgeincolour.com/tutorials/image-noise-2.htm.

I shoot landscape astrophotography and my most common setting is 20-25 sec f/2.8 at ISO 10000. I set LENR to ON but with these settings at night AUTO=ON practically. I have turned it to OFF a few times to shoot timelapses. The individual photos didn't seem different than the ones where it is set to ON but then there were so many stars visible that it wasn't easy to spot a few hot pixels. The read noise reference is interesting.

FYI, the 41% increase from subtracting a single dark frame is just math - noise increases as √2 = 1.41. Try shooting a manual dark frame (same exposure/ISO but with the lens cap on), no LENR. Open it in your favorite RAW converter, convert to JPG and check for hot pixels (whether you see them on the RAW converter depends on the converter - for example, DxO only removes them when you view at 100%, but they're gone in the JPG automatically).

Thanks. I recall ACR doing that too.

 

[niels123]

Hi Neils.
My understanding is that the dark frames are only specific to the temperature of the camera at the time of the shoot, I think that the noise is relative to the sensor temperature.

Cheers, Graham.

For *good* results, how many dark frames would you average out for 100, 200, 400, 800 and 1600 iso?

Are the dark frames specific to an individual camera or can they be used for all cameras that are the same type (e.g. 5D III).

Hi Graham,

Luckily the sensortemp can be found in the exif ;D
So with some cooling or allowing the sensor to warm up by keeping on live view, it is not so difficult to get the same sensortemp again

If I have a darkframe from a 5D III at (e.g. 22C) can I then use it on every 5D III at 22C?

 

[Valvebounce]

Hi Niels.
Got it, seems reasonable, honestly I've no idea, but I bet it won't work due to variations in the manufacturing of the mass produced items that make up the noise chain.

Cheers, Graham.

Hi Neils.
My understanding is that the dark frames are only specific to the temperature of the camera at the time of the shoot, I think that the noise is relative to the sensor temperature.

Cheers, Graham.

For *good* results, how many dark frames would you average out for 100, 200, 400, 800 and 1600 iso?

Are the dark frames specific to an individual camera or can they be used for all cameras that are the same type (e.g. 5D III).

Hi Graham,

Luckily the sensortemp can be found in the exif ;D
So with some cooling or allowing the sensor to warm up by keeping on live view, it is not so difficult to get the same sensortemp again

If I have a darkframe from a 5D III at (e.g. 22C) can I then use it on every 5D III at 22C?

 

[niels123]

Just by looking at a CR2 file or its metadata, is it possible to see if the in-camera long exposure noise reduction was set to 'on' or 'off'?

By using DPP you can know just the setting of your camera which is usefull only when it is off or on. If it is auto (the default I believe) then I do not know.

A late reply
Where in DPP can I find this setting? I've searched for a while, but cannot find it