Revisiting Color Negative Scanning and Color Spaces

 Introduction

My recent set of 4x5 color images have shown up some inattention I have had in my color scanning workflow. I first noticed it when I had a print made from Peak Imaging. The print looked noticeably darker with more muted color than the screen image. 

I had asked the lab not to correct the image where a technician will try and remove color casts and get good contrast and color balance. I wanted to try and have more control over the print. (The lab corrected images in the past I have been pleased with.)

The next thing I noticed was the same image posted to this blog and to Flickr did not reflect the same quality I saw on the same monitor when rendered on the PC. In other words the image rendered through a browser appeared differently than the image when rendered on the PC either in the photo viewing application or Photoshop Elements etc.  

Here is an example of the the screen grab of two images that are in all ways identical expect one. The image on the right is the one I originally uploaded. The one on the left better represents what I expect to see. 

“Image 2 (2) cp small sRGB” and "Image 2 (2) cp small " from Flickr screen grab

Both images as screenshot using photo viewer. Images are indistinguishable.

To save all the suspense I will say that I concluded that the workflow I used did not convert the image to have an sRGB profile. 

sRGB Origins

sRGB was developed my Microsoft and HP in the 1990s as they grappled with the advent of color images and displays and printers and how to make sure they displayed in as consistent way as possible. sRGB defines a color space, that is a range of colors that CRT monitors at the time could represent. The range of colors is referred to as the gamut. Now gamut can be confused with another aspect of sRGB that of the transfer function or gamma.  

Gamut

This term refers the range of colors that can be represented. Not surprisingly most sensors and displays cannot represent the range the human eye can. The diagram below shows the outline of human visual capability. Within that are the sRGB and AdobeRGB color spaces. They are much smaller than the range of visual capability because of the technology they are meant to represent. In the case of sRGB the range of CRT phosphor colors (and the prevalent 8-bit technology of the day). 

 

CIE Chromaticity with sRGB and AdobeRGB color spaces. 

Gamma

Gamma is a transform that is applied to an image to correct for non linear response in the vision or the technology used to display the image. We are familiar with this in digital image manipulation when we change a mid-point on a histogram or manipulate the light curves in Photoshop to increase contrast or lighten or darken highlights or shadows. Without a proper gamma an image will appear flat and lackluster.

So we now understand that whatever the limitations of sRGB it is important in that it allows images shown on screens and the web to stand half a chance be being seen as intended. Normally this all works behind the scenes as digital cameras and phones all use this and all displays use this as well. So when you take a photo on your smartphone to send to Instagram everyone see a reasonable rendition of the photo you took.  It is not perfect as the sensors and screens are not well calibrated in most cases as this is too expensive a process to use in manufacturing. I know for instance I have an old LCD monitor and a newer one and there is a definite color cast to the older one. 

Workflow Implications

For me I had to look into the workflow I used to create the image and how I got to here. 

I scanned this negative using Silverfast for this session. I wanted to try out the color conversion and dust removal. Normally I use Vuescan having been a long-time user. 

Part of my experimenting was to get a 48bit 'raw' TIFF file for use in my ColorPerfect plugin I use in Photoshop Elements 10 (yes I know old outdated SW. I am still trying to keep my SW expenditures to a minimum.) 

I outlined my first attempt at a workflow for color in this post from July 2018. The ColorNeg portion of ColorPerfect works best with a 16-bit image that has no gamma or colorspace applied. When I used Vuescan then a profile was not placed by Vuescan with 'raw' files. Elements would assume sRGB so it came automatically when the files was saved. Thus I never had the problem with older scans. 

SilverFast however in the way I originally used it applied an Epson ICC profile. This in turn was retained through the Elements flow when saved as a JPEG file. A blessing and a curse. A curse because this was the source of the problem (not saving an sRGB profile). A blessing in some sense in that this gave me what I felt was a better final image. Indeed when I went through the 'correct flow' of removing the Epson ICC profile I got an image that was close but I was not able to duplicate the result despite substantial effort. The red in this 'proper' version was not as strong and the birch tree bark had a slight pinkish cast that I did not like as much. 

I made some square test images to be sent to the color lab for printing. The two examples below illustrate the differences between the scan that started with the Epson ICC profile and the one without. 

Epson Color Space converted to sRGB

No Scanner Profile sRGB applied in Elements

Revisiting Vuescan

This work caused me to look again at Vuescan 'raw' scans. I made a new scan of the same negative with Vuescan in 'raw' format. It has no default profile applied as expected. When imported into ColorNeg and the same corrections applied the image is virtually identical to the one processed the same way with a SilverFast scan. Interesting however that the two 'raw' TIFF files have much different appearances. This because when Vuescan is setup as a color negative it attempts to remove some of the orange mask from the image. 

Raw comparison SilverFast-Vuescan. Vuescan set to 'Color Negative' 

If it is setup to treat it like 'Slide Film' the scans look the same.

Printed Output Test

As mentioned above I made some test images that I cropped square. I submitted these to Peak Imaging to print as satin matte C-Prints (Fuji photo paper). I printed them as 5x5 inch prints with small borders and text of the filenames to help identify them, printed in the borders. I order them all as 'uncorrected' so I was sure the lab was not changing the images. Peak Imaging wants all images to be submitted as sRGB color space. (This makes it simpler as I don't have to juggle two color spaces.) 

Below is a composite image of scans. 

Test Prints
 (Clockwise starting in upper right corner. 1. Original Epson Color Space, 2. Epson to sRGB Conversion, 3. sRGB from start, 4. Epson to AdobeRGB)

The differences are apparent in the prints. My favorite is still the Epson to sRGB conversion owing to the lack of color cast in the tree bark and redder color in the dead bracken. The prints however look like they could use some brightening. My next set of experiments may be to run different brightness adjustments as well as have the lab 'correct' one to see the result.