Allow multisampling with ME

[degrub]

Using the Nikon CS5000 i am not able to run multisampling when MultiExposure is on ( did i miss a switch ?). Other than intermediate storage size ( which would only be a factor of 4X worst case, 2X for each exposure ) this restriction does not make sense to me. The purpose of MS is to reduce random noise in the dark areas, ME to increase the dynamic range. Why create a noisy expanded dynamic range image ?

i request that this restriction be reconsidered.

[LSI_Morales]

Dear degrub,

Using the Nikon CS5000 i am not able to run multisampling when MultiExposure is on ( did i miss a switch ?).

No, it is not possible now and was not possible since version 6.5 (and I believe even previous versions of SilverFast too)

The purpose of MS is to reduce random noise in the dark areas, ME to increase the dynamic range. Why create a noisy expanded dynamic range image ?

You are on the right track but are confusing things.

The purpose of MS is to reduce random noise by doing multiple scans and then calculating the average tone or detail. (noise does not appear always in the same spots) This approach helps reduce noise not only in the dark areas but accross the entire range of tones.

Apparently it works best on the dark areas because digital CCDs or CMOS are prone to produce "electronic noise" in low light situation.

Multi-Exposure on the other hand, makes 2 scans. The first one at a normal scanning speed. The second is a lot slower (some times up to 4 x slower) overexposing the entire image.
By overexposing the picture, the dark areas will be up to 4 times lighter reducing the possibility of electronic noise from happening.

As you can see Multi-Exposure not only expands the dynamic range but it also reduces noise as it happens with Multi-Sampling.

You might be wondering. Then why having both?

Well. Multi-Sampling can be implemented in all scanners because is a software based solution.

Multi-Exposure is completely hardware dependent and only available for some scanners which allow the speed control of their mechanics.

As you can see they are different approaches to the same problem with the added advantage of Multi-Exposure being able to not only reduce noise on the dark areas but also capture a broader dynamic range.

Cheers

[degrub]

By overexposing the picture, the dark areas will be up to 4 times lighter reducing the possibility of electronic noise from happening.

Sorry, but i believe this is not technically correct. The noise does not go away. The signal to noise ratio does change which makes the apparent noise level diminish. Perhaps that is what you were referring to ? Thus there would still be advantage in multi-sampling during each of the two passes. Perhaps because the S/N ratio is higher , i would only need to take 2 or 4 passes instead of 8 or 16. This is what i am requesting.

Nikon also did this when they chopped off the bottom 10-15 levels. Using their "analog gain" allowed one to shift the levels to the right as you mention, but it did not reduce the noise, it only increased the signal to noise ratio. And i could still use the multisampling feature to reduce the actual random noise.

As you can see Multi-Exposure not only expands the dynamic range but it also reduces noise as it happens with Multi-Sampling.

Only to the extent that the two passes gets about 1/2 bit of dynamic range worth of noise reduction.

Please consider removing the restriction of not being able to use multi-sampling when using ME.

Thank you.

Regards,
Frank

[LSI_Muenier]

Dear degrub,

both methods (MultiSampling and MultiExposure) increase the signal to noise ratio (i.e. reduce the noise).

Theoretically, MultiSampling reduces noise by a factor of 1/squareroot(n) where n is the number of scans combined.
To achieve a reduction to 1/4th you need to use 16 (!) passes, roughly.

MultiExposure on the other hand reduces noise by a factor of 1/4th or 1/3th depending on the capabilities of the scanner.
So, this is comparable to the reduction of 1/4th by 16-pass MultiSampling (which is the maximum in SilverFast).

Currently, SilverFast does not allow to combine the two methods, although this of course could be possible in principle.
We have not yet sorted out if the much longer scanning time (remember: you need to multiply scan at high exposure)
really is worth the result.
And don't forget that the improvement by MultiSampling is only proportional to the squareroot of the number of scans combined.

Best regards
Martin

[degrub]

Understand.

What i am asking for is the software to allow the user to make that choice and decide if the additional scan time is appropriate for the image.

Regards,

Frank

[LSI_Muenier]

Dear Frank,

I understood that, but we simply have not implemented this feature into SilverFast.
There is currently no schedule for this, but I have recorded your interest in this.

Best regards
Martin