73 megapixel scan of a 6x9 negative (Portra 800)

[cerber0s]

Kodak Portra 800 shot on a Fujica Professional 6X9 and developed in Cinestill 2-bath developer.

My scanning setup is far from perfect, but it gives an idea of the "resolution" in these negatives. Click image for full resolution.

 

[MattKing]

Moved to the Hybrid neighbourhood.
Thanks for sharing this.

 

[Mick Fagan]

The closest exhaust pipe, looks like a snake about to bite; the curvature of the kerb along with the highlight spots that look like eyes, just make this very interesting possibility for an abstract picture.

If you zoom in as much as the software allows, then you can see the almost facial impression of a snake face!

 

[runswithsizzers]

It is hard for me to interpret what I am seeing online without knowing the actual magnification (zoom level). I can't be sure, but I think the browser software used for viewing your photo online is capable of over-zooming -- that is, I believe your "full resolution" is being magnified to some resolution which is greater than 100%.

When I download the image and set Adobe Photoshop to view the image at 100%, it looks much better -- and displays as much smaller -- than it does when viewed online at the maximum zoom level. I have to set Photoshop's display level to about 200% to make it the same size as the maximum size when viewed online.

From experience, I know what my own scanned images look like at actual size (100%), and I know they look much worse at 200%. This is a problem when comparing sharpness and resolution of images online. How do we control / define what size the images will be viewed at? (I don't have a good answer.)

 

[koraks]

How do we control / define what size the images will be viewed at? (I don't have a good answer.)

The preferable option in my personal (non-moderator) opinion is to do the following:
1: Post a small version of the entire image, resized to a convenient size so that it fits in a post. I generally use 800-1000px for the longest dimension.
2: Add one or more 100% crops to illustrate fine detail rendering of relevant area(s), again ensuring these crops are conveniently small in absolute size.

In this particular case, this might look like this:

Entire image

Crop 1

Crop 2

Coincidentally, and not to rain on anyone's parade, but crop #2 illustrates how the colors are pretty badly mangled in some places. There's also a great deal of color fringing (blue/yellow) that I suspect originates in digitization of the negative, not in the original exposure. Of course, we're pixel peeping here - which was kind of the point of the exercise, so perhaps it's OK if I note these issues.

 

[cerber0s]

so perhaps it's OK if I note these issues.

Most definitely!

 

[cerber0s]

colors are pretty badly mangled in some places

I’ve rescanned and converted this multiple times, and noticed that the reds are unruly every time. Other scans have also come out with certain colors looking over saturated.

I haven’t figured out what causes the problem yet.

 

[loccdor]

I’ve rescanned and converted this multiple times, and noticed that the reds are unruly every time. Other scans have also come out with certain colors looking over saturated.

I haven’t figured out what causes the problem yet.

It looks to me like what happens when colors are out of gamut. I don't know your process but sometimes changing the ICC profile from sRGB to Adobe1998 helps with that. The exposure level of the scan could also be playing a role.

 

[runswithsizzers]

@koraks - yes, the method you describe seems to be the best solution for controlling / describing what a 100% crop looks like. As a test, I just tried to (more-or-less) duplicate one of your 100% crops using Photoshop. Looks like we got about the same results -- so a crop from an image at 100% magnification can be used a point of reference for comparison purposes. When exporting a crop from our various editing applications, we may each be using somewhat different output sharpening settings, JPEG compression, etc, which will introduce some variation, but any minor variation due to export settings is probably going to be much less of an influence compared to the difference between viewing at 100% vs. 200%.

 

[koraks]

I just tried to (more-or-less) duplicate one of your 100% crops using Photoshop.

Yeah, that's essentially what I did, but I used GIMP. Same principle though!

When exporting a crop from our various editing applications

Absolutely. I simply made a screen shot using the Windows snippet tool and pasted it into my post. It's the quickest way. IDK what kind of compression is added. No sharpening that I know of.

It looks to me like what happens when colors are out of gamut.

Certainly so, but where in the process this happens, is not clear to me yet. It might just as well happen during the inversion & color balancing step.

I haven’t figured out what causes the problem yet.

Yeah, these issues can be difficult to troubleshoot. I'd start by looking at the histogram of the original capture, then keep an eye on it as you invert, color balance etc. At some point you should start seeing data on the red channel bunching up towards the extreme edge of the histogram. But like @loccdor says, an ICC conversion step can also cause this and those can sometimes be tricky to identify.

 

[cerber0s]

It looks to me like what happens when colors are out of gamut. I don't know your process but sometimes changing the ICC profile from sRGB to Adobe1998 helps with that. The exposure level of the scan could also be playing a role.

This was a good tip, it seemed to tame the unruly reds. Thank you!

 

[cerber0s]

I uploaded the original raw file, if anyone wants to play with it.

 

[koraks]

I uploaded the original raw file, if anyone wants to play with it.

I have it open, but the highlights look really funny. Are you sure the film is properly fixed?
Also, the blue channel seems to be clipped in the digital file; I'm running out of steam on the blue highlights - they blow out:

I think you may need to give a little bit more exposure when digitizing the negatives so you end up with more headroom on the blue channel. I think it'll make color balancing the images easier, the slope-off into the specular highlights will become more smooth and the overall risk of ending up with out-of-gamut colors is lower.
Here's the RGB histogram in my RAW converter:

Note how the blue channel is all bunched up to the left. You've got some room on the right on both R and G, so I'd suggest using that in order to give blue a little more room. Dial in 2/3 stop more exposure, see how that goes.

I also have to admit that I have some doubts about the Cinestil CS41 developer used. The colors are just...wonky. I can't get them to balance out, and what I need to do to get close to OK, is very non-linear:

What certainly doesn't help is the uneven illumination of the negative, that throws the color balance into different directions across the frame. Look at e.g. the sunlit tarmac along the left edge of the frame vs. the center of the frame above the saddle of the bike. Here's a sample of both areas:

I balanced the light-midtones to give a more or less neutral hue on the tarmac, but then it goes lavender in the left and right of the image.

All this doesn't show up very clearly in the original image posted in #1, but that's mostly because it's very desaturated. Crank up the saturation and the same problem pops up, so it's not something to do with my RAW converter etc. (Also note that Portra 800 isn't all that desaturated if you optically print it onto RA4 paper. It renders quite a rich palette in my experience, so I wouldn't consider it a 'flat' film or something along those lines.)

There's a lot going on here and I think there's some room for improvement.

 

[brbo]

 

[koraks]

That's really good @brbo. How'd you approach it?

Oh, I think I got the gist of it. It's an adjustment I never find necessary on color scans, but here it saves the day. It's along these lines:

 

[brbo]

I didn't have to do anything special, really. As always I used libraw (dcraw) to process the raw file to get linear tiff file. I then used my standard routines for inverting and nothing seemed to be off.

I now used Adobe raw processor and still the inversion is... uneventful. Not surprising, since the histogram of the image area looks pretty decent.

 

[koraks]

It may have something to do with how Rawtherapee handled the raw on my end. I had to desaturate the blue channel dramatically in post to get a similar result to yours, which is something I never need to do on my scans. I suppose Rawtherapee somehow adjusted the white balance in such a way as to compress the blue channel; see the histogram it showed. It's very different from yours.

The color balance variation across the frame is present in your version as well, so that's really part of the capture.

 

[cerber0s]

Are you sure the film is properly fixed?

Well... I'm sure the temperature and times were correct. This was the second roll I developed in a freshly mixed batch of chemicals, and I compensated for it not being the first. I can't be sure nothing else is off though.

If I had known the thread would take this turn I would have told you that the film was one month expired when I bought it a year ago. I've kept it in the freezer ever since. I do however appreciate the turn this took, as I keep learning stuff from you guys.

Your version looks really good @brbo

We're also dealing with at least three different softwares, and three different methods of converting I think.

@koraks my histogram looks more similar to yours.

After I set the white balance to the film base and select Adobe RGB profile:

 

[cerber0s]

I then crop, because I don't want the film base in the histogram data, and adjust each individual RGB curve to just before they clip:

I then do fine adjustments to the curves using the RGB parade to gauge how even they are:

Then I I leave RAW developer and add an Invertion layer

Then I use the curves to fine tune again:

I think it's incredibly difficult to get the colors "right".

 

[koraks]

I don't think the film being expired is much of a factor, really. Much of the color balance issue I noted traces back to how I processed the raw file.

In terms of final color balance it's of course always a matter of preference. The possibilities are infitine! I quite like @brbo's result, but that's generally the case looking at his color scans and prints.

 

[cerber0s]

I quite like @brbo's result

So do I.

I don't think the film being expired is much of a factor, really.

Neither do I, but it was worth mentioning.

 

[mshchem]

This is all pretty mind boggling. Scanning (and printing) color negatives is quite an art!

 

[cerber0s]

I then used my standard routines for inverting

Inquiring minds want to know

 

[runswithsizzers]

Here is a quick go at it using the Negative Lab Pro plugin for Adobe Lightroom. Normally, after making the basic conversion to color with NLP, I would work on it some more in Lightroom. Specifically, I would probably select the upper right quadrent with a linear gradient and try to warm it up a bit to better match the color temperature the rest of the scene. And those blue traffic signs might need a bit of taming as well. I can post the Negative Lab Pro settings I used if anybody wants to see them.

Edit: Compared to @brbo's version, it looks like I got my color balance a little warm -- except for that right side which is too cool.

 

[cerber0s]

Here is a quick go at it using the Negative Lab Pro plugin for Adobe Lightroom. Normally, after making the basic conversion to color with NLP, I would work on it some more in Lightroom. Specifically, I would probably select the upper right quadrent with a linear gradient and try to warm it up a bit to better match the color temperature the rest of the scene. And those blue traffic signs might need a bit of taming as well. I can post the Negative Lab Pro settings I used if anybody wants to see them.

View attachment 376176

That looks good too. The traffic signs actually look close to real life.