Scanning colour in Vuescan: "RAW tiff" or "RAW DNG"?

[DillonVFX]

When scanning colour negative film in Vuescan should I have Vuescan output the file as a "RAW tiff" or as a "RAW DNG" when I'm going to be bringing the photos into Lightroom and Photoshop for processing with ColorPerfect? Currently I'm following this guys video and he says to output the files from Vuescan as a tiff but is this really the best way to preserve the film latitude and have the most flexibility when processing the photos through Lightroom & Photoshop together?

 

[craygc]

Vuescan raw is not really raw in the sense that its already a demosaic image with a gamma of 1. So the adjustments you usually have in a true raw image such as white balance, colour space, exposure, shadow & highlight recovery aren't going to be there as its already baked the image and data has been lost.

I've never found any benefits from scanning in raw (tif) with Vuescan (Nikon scanners) that don't out way the latter issues of gamma correction

 

[Godfrey]

Scanners are not RGB capture sensors, so there is no demosaic going on.

The TIFF and JPEG output from a VueScan capture has had image processing applied, depending on the parameters you set up in the various settings panels. This includes image cleaning with dust removal, sharpening, gamma curves and brightness adjustments, etc etc.

The 'raw' output image from a VueScan capture is a linear gamma RGB TIFF file with no inversion (for negative films) or other image processing applied. Bit depth, etc, are all dependent upon the particular scanner being used along with the capture settings.

When you tell VueScan to output the raw file as DNG, you are getting the same raw scanner data wrapped into a DNG container file. With a DNG file, you can utilize the Adobe tools to develop customized calibration profiles, etc, for processing of the DNG files to your preference.

I do the DNG workflow as it works well for my needs.

G

 

[Pete B]

I thought it needs to be the Tiff Raw if it's to be processed using ColorPerfect?
Pete

 

[willie_901]

This.

Scanners are not RGB capture sensors, so there is no demosaic going on.

The TIFF and JPEG output from a VueScan capture has had image processing applied, depending on the parameters you set up in the various settings panels. This includes image cleaning with dust removal, sharpening, gamma curves and brightness adjustments, etc etc.

The 'raw' output image from a VueScan capture is a linear gamma RGB TIFF file with no inversion (for negative films) or other image processing applied. Bit depth, etc, are all dependent upon the particular scanner being used along with the capture settings.

When you tell VueScan to output the raw file as DNG, you are getting the same raw scanner data wrapped into a DNG container file. With a DNG file, you can utilize the Adobe tools to develop customized calibration profiles, etc, for processing of the DNG files to your preference.

I do the DNG workflow as it works well for my needs.

G

In VueScan raw just means an original, unmodified image. In digital photography raw means a data array of RGB values that require demosaicing to render an image (except for a couple of cameras with out color-filter arrays).

So the term raw is mis-used in VueScan now that digital raw capture is ubiquitous. Unfortunately this is creates confusion.

 

[Godfrey]

This.

In VueScan raw just means an original, unmodified image. In digital photography raw means a data array of RGB values that require demosaicing to render an image (except for a couple of cameras with out color-filter arrays).

So the term raw is mis-used in VueScan now that digital raw capture is ubiquitous. Unfortunately this is creates confusion.

VueScan raw means the data from the capture device, a scanner, without processing applied. Scanners are different from digital sensors in cameras, that's all. Nothing confusing about it, to me anyway. VueScan was around and defined its raw data format before raw digital files from cameras were all that common.

G

 

[Ranchu]

VueScan raw means the data from the capture device, a scanner, without processing applied. Scanners are different from digital sensors in cameras, that's all. Nothing confusing about it, to me anyway.

So, how is a scanner sensor different than a camera sensor? If it's not RGB filters on a CCD, how does it see color?

 

[Godfrey]

So, how is a scanner sensor different than a camera sensor? If it's not RGB filters on a CCD, how does it see color?

A typical camera sensor has a 2D array of photosites with a Bayer mosaic of R, G, and B filters over the photosite lenses. Thus the term "demosaic" means to obtain the color mix for each destination pixel in the target 2D RGB space by interpolating each pixels best-match probable RGB mix from the surrounding box of nine photosites. What it's doing, essentially, is creating the full color image by creating three 2D arrays, one in R, one in G, and one in B, where each position (x,y) becomes the sum of (xR,yR), (xG, yG), and (xB, yB).

Flatbed and film scanners have a variety of sensor designs, the simplest might be a single, moving strip of three photosite rows, each row being monochromatic R, G, and B. The scanner driver in that sort of design samples the subject linearly and directly assembles a set of R, G, and B 2D arrays, which add together in the same manner as the Bayer mosaic sensor after demosaicing, to assemble the complete color pixel map.

There are other designs, that's just a one simple type for illustrative purposes. But none use the Bayer mosaic design that's used in camera imager sensors to my knowledge.

G

 

[Ranchu]

Great, that's what I thought.

 

[anerjee]

Does this mean that resolution numbers are not comparable between scanners and cameras?

In other words, if I scan at 3000x2000, it must mean I need a more than 6MP camera to match the same resolution? Assuming the film is sharp enough and the scanner is also resolving at that level.

A typical camera sensor has a 2D array of photosites with a Bayer mosaic of R, G, and B filters over the photosite lenses. Thus the term "demosaic" means to obtain the color mix for each destination pixel in the target 2D RGB space by interpolating each pixels best-match probable RGB mix from the surrounding box of nine photosites. What it's doing, essentially, is creating the full color image by creating three 2D arrays, one in R, one in G, and one in B, where each position (x,y) becomes the sum of (xR,yR), (xG, yG), and (xB, yB).

Flatbed and film scanners have a variety of sensor designs, the simplest might be a single, moving strip of three photosite rows, each row being monochromatic R, G, and B. The scanner driver in that sort of design samples the subject linearly and directly assembles a set of R, G, and B 2D arrays, which add together in the same manner as the Bayer mosaic sensor after demosaicing, to assemble the complete color pixel map.

There are other designs, that's just a one simple type for illustrative purposes. But none use the Bayer mosaic design that's used in camera imager sensors to my knowledge.

G

 

[Godfrey]

Does this mean that resolution numbers are not comparable between scanners and cameras?

In other words, if I scan at 3000x2000, it must mean I need a more than 6MP camera to match the same resolution? Assuming the film is sharp enough and the scanner is also resolving at that level.

It depends on how you want to measure resolution. In a Bayer matrix type camera sensor, the spatial resolution in pixels is fixed to the dimensions of the 2D array of photosites minus the edge photosites used to interpolate color values. Color values are interpolated for every pixel in the output array.

Scanner resolution for the type of scanner I used as illustration above is the number of rows of sensing elements in one dimension by how fine a pitch the stepper motor can stop the sensor assembly. If that evaluates to the same number of 2D pixels in the output array as what the camera does, the spatial resolution is the same, but since the three rows are collecting color information for each of those pixel locations, the color information is three times as much.

So the scanner in that example would have the same 6Mpixel spatial resolution, given a 3000x2000 output matrix, but perhaps finer gradation of color tones depending upon the quality of the scanning elements vs the quality of the Bayer demosaic algorithm.

All that is given the ideal case of considering the image to be perfectly focused for the scanner or camera sensor, and without considering the interactions of the film medium (base, emulsion grain, etc) on the image itself. Digital camera capture has little of those interactions to consider.

So, in my opinion, it is very difficult to compare digital camera capture with cameras vs scanning resolution with any degree of credibility or practical value. It's a remarkably slippery thing.

G

 

[anerjee]

I don't understand the details, but your conclusion seems quite reasonable.

I cannot find a lot of detail difference between my Pakon Scanner at 6MP vs. Sony A7S 12MP image when I use a ISO 100 fine grain film with a super sharp lens like the 45mm Planar G.

The digital allows me shoot at that level of sharpness at ISO 40,000 🙂

It depends on how you want to measure resolution. In a Bayer matrix type camera sensor, the spatial resolution in pixels is fixed to the dimensions of the 2D array of photosites minus the edge photosites used to interpolate color values. Color values are interpolated for every pixel in the output array.

Scanner resolution for the type of scanner I used as illustration above is the number of rows of sensing elements in one dimension by how fine a pitch the stepper motor can stop the sensor assembly. If that evaluates to the same number of 2D pixels in the output array as what the camera does, the spatial resolution is the same, but since the three rows are collecting color information for each of those pixel locations, the color information is three times as much.

So the scanner in that example would have the same 6Mpixel spatial resolution, given a 3000x2000 output matrix, but perhaps finer gradation of color tones depending upon the quality of the scanning elements vs the quality of the Bayer demosaic algorithm.

All that is given the ideal case of considering the image to be perfectly focused for the scanner or camera sensor, and without considering the interactions of the film medium (base, emulsion grain, etc) on the image itself. Digital camera capture has little of those interactions to consider.

So, in my opinion, it is very difficult to compare digital camera capture with cameras vs scanning resolution with any degree of credibility or practical value. It's a remarkably slippery thing.

G