Scanning negatives with DSLR - need help with setup

rory_clague · April 5, 2015

One more thing, as noted above, be sure to use a remote shutter release and mirror lock-up or live view. I always re-focus for every frame I take of the negative, as well.

As the images above didn't seem to post, here is another try:

etienne_w · April 22, 2015

Rory, thank you. This is awesome stuff. It is pretty much the workflow I use too, although you have developed its details

much further than I have. So if I understand you correctly you set the curves individually for each negative you scan?

With positive slides I generally use the same setting for all. I bracket at the scanning stage and select the best exposed

picture (optimal exposure might vary with the density of the slide). I apply little, if any, post processing after that. I

generally leave the curves alone. I have been searching for a one-size-fits-all kind of setting for negatives but without

success so far. This might simply not exist as you tend to imply.

Happy shooting!

Etienne

rory_clague · April 25, 2015

Hi Etienne, sorry for all the details! I kind of want to use my own post as a guide for myself and for others who might want to try the method. I do set the curves individually for each negative. I tried making a preset a couple of times, and I find that this just never works to my satisfaction. I imagine variables in development, density, and other issues change the relationships between how a digital camera "sees" the negatives and therefore what correction factors should be applied. I do make some 'darkroom' choices in this process that aren't strictly the same as what a scanner would choose, or what an optical printer of a color negative might do. Of course, the only reason we regard a scanner's choices as somehow privileged is that the scanner's software is using a preset, albeit a very good general preset that the scanner and software engineers worked very hard to optimize. I simply find that doing my own optimization for each image works out better for me. It takes a bit of time, but not all that much, and for me it's part of the fun of using a hybrid process. If I were using an enlarger, I would have fun playing around with the prints I could get by making choices there, too (and I do definitely plan on getting a color enlarger someday).

Anyway, to blabber on some more, I'll just share my processing workflow right here for the record. Here, I'll just show the process I chose for an unremarkable shot I took of a friend with 35mm portra 160. From my DSLR scan, I had 10 digital raw images of the 35mm negative. With a 5D3 or D800 series, I could have fit the negative in one image at 1:1, of course. I like to sample as much resolution as possible, which is why I put as many pixels as possible onto the negative. I've found since I did this that 6 or even 4 shots is enough for a negative with APS-C. I did way too much overlap early on to try to optimize the stitching process with Microsoft ICE. So, to start, I choose an image that has a good view of my main subject, or an image that has a wide variety of colors and densities. All of my processing is done in DxO Optics Pro, but other raw converters have the same tools and the same workflow will apply in Lightroom, ACR, and other converters.

The first thing I do is account for small defects in the 7D/100L combination used for digitizing the negative. So, that means in DxO I turn on noise reduction, distortion correction, vignetting correction, chromatic aberration correction, and DxO lens softness correction, which applies sharpening to a raw file in a way calibrated to your particular lens. I usually turn up the sharpening very slightly form the default settings so that the grain looks sharp. Of course, you don't have to do this. I want the grain sharp, but not harsh, so I also typically turn down the microcontrast to -30 or so. Since the pixel density with this approach is quite high, the overall microcontrast of the image isn't negatively affected too badly. It just renders the grain a bit less harshly. See settings on the right panel. Again, other software packages like LR have these corrections.

Next, I choose a white balance that I will use for all of the raw images in the collection. I choose this by using the white balance selector to click on a few different "white" (or "clear") pixels in on the edge of the negative. This compensates for the brown-red color of the film backing. I click on a few different locations in this white area and select a Kelvin temperature and tint that is roughly the average of all these clicks. In the case of this image, it was roughly 3100 K and -20:

I then select all the images, and apply the same white balance settings to all the images. It's important for this workflow to do this before manipulating the R, G, and B curves. <img src="https://farm9.staticflickr.com/8742/17080143370_9061e56c3f_o_d.png" alt="" />

Next, I go back to the image with the best view of my main subject. I then look at only the red histogram and turn on the red curve. Note how the red histogram shows all the pixels in a small region of the histogram. What I want to do is map those pixels with the red curves tool so that they are spread out over the entire luminance range.

The first step is to simply invert the red curve, like this:

Now the red highlights in the negative are now in the shadows in the positive. The histogram, though, still looks compressed in one small region. Now I use the curves tool to remap where the white and black points are for the red curve. After this, the histogram for red should be distributed over the entire luminance range. Be sure not to extend it too far. You want the brightest red pixel to be remapped to 255,0,0 and the darkest to be remapped to 0,0,0, but you don't want the second-brightest red pixel to take on this value instead, or you will get blown highlights and clipped shadows. Now you can see that the red curve has a very distinct shape:

Next I do this same process with the green channel. First I invert the green curve:

Then use the green histogram as a guide to remap the tonal values of the green channel over the entire luminance range using the green curves tool:

Then I do the same thing with blue. First invert the blue curve:

Then remap so that the histogram is distributed across the luminance range. Interestingly, you can see the blue channel requires less remapping than the green channel, which requires less remapping than the red channel that was originally quite compressed in the negative:

Now that the white and black points are set for each curve, I usually introduce a slight bulge in the middle of each curve, along the axis of the original white luminance (or RGB) curve. This brightens the midtones slightly and, depending on the magnitude of the changes, applies slightly different weights to each of the R, G, and B curves. This is where aesthetic choices become somewhat significant. Usually the red channel has the strongest curve, followed by green, followed by blue. Sometimes, the blue channel even gets a negative curve. In this case, the blue channel was left mostly flat. Finally, I will sometimes modify the combined RGB curve to make the exposure what I want. These steps are not necessary, but I find they always improves the look of the image:

Finally, in DXO, I copy the correction settings (which includes the curves manipulations) and paste them to all the images that will be stitched to form an image:

Then export as 16-bit TIF files. I've tried exporting as DNG files, but somehow Microsoft ICE doesn't read the curves functions I applied and the resulting colors are a real mess. Therefore, I stick with TIF files. If you aren't as crazy as I am, you could just use jpegs, too:

Finally, I use Microsoft ICE and drop all the TIF files, choose planar with skew, crop out the black frame around the edges of the image, export as a TIF file, re-open the TIF in DxO, and use the clone stamp tool to deal with dust. Then I'm done. It sounds like a lot of work, but it is really quite fast once you get the hang of it.

Finally, you have a finished image that looks okay. Black and white is very simple compared to this. Just convert to greyscale and invert the combined RGB curve rather than the R, G, and B curves individually. Introduce an S curve or not depending on preferences in order to get some more contrast. Since black and white film has a higher density range than color negative film, remapping the tonal values isn't usually necessary. Same goes for positive film (which is the easiest of all - I don't change the curves at all with color positive film).

harry_hoag · April 26, 2015

Wow, nice Rory, great to see your workflow thanks for that, and great results. I tried DSLR scanning with my 550d a while back, but quickly realised how fiddly and unsatisfying the results were, particularly for medium format. Some of you guys here seem to have got some really great results though! Mostly out of impatience (and lack of a really decent DSLR) I opted for a flatbed negative scanner which I've been really happy with. I'm definitely going to try manually inverting all the channels after reading Rory's post, but doing it on every picture in a roll of film is painful. I normally let the scanner invert the negative for me with no other colour correction. Then I use photoshop with a really handy automation to set up layers for setting the white black and grey points with the help of the threshold effect (let me know and I'll post an explainer too if you like). I can then manually grade to preference. In terms of dust & scratch removal, I really hate the automatic ICE things, they take out stuff that aren't dust and scratches sometimes and leave ugly splotches. For me, by far the most accurate and quickest way to do it is with the photoshop spot healing brush tool, it's absolute magic! Way better than the 'traditional' stamp tool.

Sandy Vongries · January 18, 2016

All the slides posted under Nikon Color Slide Conversions in my portfolio were done with a Nikon ES-1 (which mounts to the 52mm filter ring, so an adapter could be used for other brands) a 55 micro 3.5 AI, and an M-2 ring on my DF. No tripod, I use a LED desk lamp -- easy as pie. It won't do negatives, for that I have to use the bellows / and PS 4. That setup requires the tripod, etc.

I ordered a Canoscan 9000f MK2, mostly for photos, but I will be comparing the results for slides, and will I.D. those when I post them.

Results are not perfect with the first setup, but quick and not too bad.

Sign In

Scanning negatives with DSLR - need help with setup

Recommended Posts

rory_clague

Link to comment

Share on other sites

etienne_w

Link to comment

Share on other sites

rory_clague

Link to comment

Share on other sites

harry_hoag

Link to comment

Share on other sites

Sandy Vongries

Link to comment

Share on other sites

Create an account or sign in to comment

Create an account

Sign in

Explore

Forums

Learn

Job Opportunities