Restoring Photographs


Introduction


Restoring old photographs is a very complex affair, but we have to start somewhere. Scanning old photographs also involves thinking about ways to restore tone and colour, and even repair some damage. Old photographs will have blacks faded and dingy whites, or what is called a narrow tonal range. Expanding the tonal range can bring back some natural brilliance. Shadow and highlight adjustments can do a lot to make an old photograph look fresh and new. 

Restoring colour is just as important for b&w photographs. Many old b&w photographs will now be a mix of browns and yellows. Old colour photographs almost always need colour restoration. In colour restoration hue and saturation controls are also very important. There are also tools for removing scratches, etc., and even for repairing damage such as stains, etc. Family snaps from the mid-20th C usually have grayish blacks, maybe some browns and yellows, and possibly some fine cracks and scratches. More recent photographs will have better colours, but they may still have lost some colour saturation. Photographs on display will usually be very faded. Pre-1930’s photographs will almost always be b&w, or now anything from pale tan to dark brown with low contrast. Here the question is do we keep the look of the old photograph as it is now, or to try to return its original look and recover the deep blacks and bright whites? Slides usually keep their colours well and will not require much tone or colour restoration (although non-Kodachrome slides might require more work). 


On this Webpage I will use Pixelmator and the book 'Digital Restoration' by Ctein to learn to scan, restore, and archive photographs taken from my collection. 


Ctein hit the nail on the head when he wrote that things might look simple. You scan the original, use an image processing program to correct the defects, print the finished photograph, and archive the restored image’s digital file. The problem is that it glosses over the real work that’s hidden in the three magic words ‘correct the defects’. It was just these three words that condemned HP’s Scan software. It was impossible for me to understand how their software ‘corrected the defects'. What did their corrections do, how did they work and how could a series of on/off buttons do justice to my archive of photographs?


Again Ctein gets it right when he notes that the key is to do just enough restoration to get the job done. In some cases this could mean just correcting a slight overexposure, but in other cases it could mean improving the colour and tones in an old, faded photograph from the 50’s. How can software such as HP Scan do that with its all-or-nothing approach with on/off buttons?


Ctein has, at times, a perfectionist view of photograph restoration. However he is right to highlight that a good photograph (and restoration) has:

Rich and accurate colour

Accurate tonal rendition

Good density range in the print

Good and sharp detail. 


The 'art' in photograph restoration is when it goes beyond just restoring the original photograph, and includes trying to actually improve the composition, exposure, focus, and original printing. 


I am unashamedly using Ctein’s book as a backbone for this page. He identified 5 key restoration activities, namely:

Restoring Tone - Photographs in need of restoration usually don’t have very good tonality. Fading and staining will wash out blacks and make whites dingy and dark. A severely faded photograph will have a very narrow tonal range. A big part of restoration is expanding the compressed set of tones back to its original natural brilliance.

Restoring Colour - Both B&W and colour photographs need their colour restored. Most old B&W photographs may now be brown and white, brown and yellow, or even dark brown and not-so-dark brown. Part of the restoration job is getting that photograph back to its original hue. Colour photographs (prints, slides, and negatives) almost always need colour restoration. 

Fine-Detail Repairs and Cleanup - Old photos invariably need to be cleaned up. They will be dirty and scratched and have fine cracks or crazed surfaces or annoying textures. 

Major Damage Repairs - Photographs can be torn, missing emulsion, or even in pieces. 

Repairing Uneven Damage - Photographs can have streaks and stains or be tarnished in some parts.


The majority of my photographs fall into a few often-encountered groups. Firstly, colour ‘snapshots’ taken from the mid-20th C. Colours have lost some intensity and saturation. Older prints have faded, some seriously because they have been on display. Secondly, some are earlier b&w photographs that are low in contrast, with greyish blacks, and in some cases brown or yellow where the silver image has broken down. Thirdly, we have a few school and commercial portraits. Fourthly, quite a number are digital prints, and I have still the original files. Fifth, we have slides. Lastly, we have negatives (colour and b&w). 


Restoring tone


According to Ctein, the First Rule of Good Tonality is that a good photograph almost always has a range of tones that run from near-white to near-black. If you assume that the photograph you are restoring is supposed to have a full range of tones, you will almost never be wrong. Conversely, a good photograph usually doesn’t have large areas that are pure white or solid black. Flat prints with muddy blacks and grayish whites are the most common problems. 

A good-contrast histogram spans most of the range of values from black to white. A smattering of pixels may be pure whites and blacks in the finished photograph, but ideally the histogram should have few or no pixels at the extreme limits. This is indicative of a good contrast range in the photograph.


This is the Pixelmator 'Levels' histogram for a photograph taken some 10-15 years ago. My impression is that it corresponds more or less to what Ctein requires. 


This is the Pixelmator 'Levels' histogram for a photograph taken more than 35 years ago. My impression is that it corresponds to the kind of photograph that Ctein would work to increase the tonal range during the scanning. 


This is the Pixelmator 'Levels' histogram for a smartphone photograph taken some 3-4 years ago. Here we can see some major clipping associated with overexposure.  


For Ctein when scanning a photograph he aims to get tones that fill most of the range of the histogram. He does this because it also gets him much closer to an image with good initial tonality.


The First Rule is almost an absolute law for original slides and prints made from slides. Slide film has an extremely short exposure range, and usually slides will have a true white and a true black somewhere in the scene. This is not always true for colour negatives and prints made from them. Professional portraits may lack a true white and a true black, especially if they were supposed to be soft, ‘flattering’ portraits made with diffusion filters. Amateur portraits are a different matter, and the First Rule likely applies. The old colour print processes were not as good, while films tended to be contrasty, prints were often low in saturation and lacked a really good black. Almost all colour prints from that era are severely faded and stained. 

Many old b&w photographs were not black and white to begin with, they were brown or sepia in colour. But they should be treat as though they were true b&w. Because of the staining and fading that afflict old photographs, you will normally eliminate all vestiges of hue during the restoration process. At the last stage, once the restoration is complete, you can digitally ‘tone' the print to give it the hue you want.

Moreover, most amateur photographs from the mid-20th C were terrible prints to begin with. Photofinishing back then was poor, and people’s expectations were lower. Well-preserved b&w snapshots from that era are more often grey and white rather than black and white. As part of the restoration process, grey-and-white snapshots and prints with washed-out colour can be turned into full-toned photographs.

How does Ctein get to those full-toned prints? What is needed? The whites are too dark and the blacks are very washed out. Photographs will need to be brighter and more contrasty overall. Ctein does not use the Brightness/Contrast adjustment to fix this (he says it is a tool of amateurs). Levels adjustments lack fine control and subtlety, but are useful for adjustments in the scanner software. It provides a good range of values in the scan, and the data is still pretty linear and un-massaged. The ‘Curves’ tool does everything that the Brightness/Contrast and Levels tools do, but you can also attach a multitude of control points to a curve and shape it as needed. This is the secret to getting good tonality in b&w restoration, and it’s absolutely essential for colour. Careful Curves adjustments in colour are often the only way to produce exactly the right colour corrections. 

This is the Pixelmator “Curves” tool for the above photograph taken some 10-15 years ago.


The Curves tool can do the same as a Brightness/Contrast adjustment. Raising the line in the Curves tool by (say) +32 is the same as raising Brightness/Contrast by +32. 




Making the line in Curves steeper corresponds to a higher contrast.




Our photograph from 40-odd years ago looks, as expected, faded, and at the same time dominated by the white. 


Now using Curves we can start to improve the image, by adjusting both the Brightness and Contrast by changing the black and white endpoints. Ctein mentioned that the scan software should not be adjusted to create too much Contrast, since the black and white endpoints can be adjusted later.  


With Curves it is also possible to pick a point (or in fact several points) and move the curve up or down. Up will make the middle greys lighter, and down will make then deeper. Here we can see that shadows and mid-tones have become a little deeper.



It is possible to fix several 'control' points on the curve and move each separately up and down to create a double bend (S-curve or reverse-S). We have moved the black and white endpoints nearer together. We have lightened the highlights and increased Contrast, and deepened the shadows. These images have been modified to demonstrate the effect of Curves, they are not optimised for printings. 




If we look at the below b&w photograph taken more than 55 years ago we can see that the blacks are present but offer little tonal variation, whereas the highlights are too strong and the white has been clipped.  




Summarising so far…


According to Ctein the key in any photographic restoration is to obtain snappy mid-tones which keep the final prints from having that flat, 'this is a copy' look. A really good restoration doesn’t look like a copy, it looks like an original photograph.

 

So we have seen that raising the whole line increases Brightness, and making the line steeper increases Contrast by dragging the black and white endpoints inwards (closer together).


To improve a dark and dull b&w photograph the endpoints can be moved (increasing the slope of the line) to improve Contrast. Moving the black and white endpoints nearer together can exploit better the available tonal range. Moving the endpoints should be done with caution to avoid eliminating highlight and shadow details. With colour photographs there is a difference between colours that are clipped and those that are simply saturated. The mid-point can be slightly raised above the centre point to make the photograph a bit lighter.


We have also seen that the line can be bent to form a slight curve. If the black and white end points of the line are kept unchanged then a slight bend upwards at the mid-point makes the middle greys lighter (a more 'open-looking' picture). A slight bend downwards at the mid-point darkens the middle greys and makes the photograph more somber in appearance. Such a bend or curve will also alter the Contrast in the highlights and shadows. Raising the mid-point compresses the highlights, lowering their Contrast and tonal separation. At the same time it increases shadow Contrast and makes those tones more clearly visible. Lowering the mid-point increases highlight Contrast at the expense of shadow separation.


For many photographs there is a lot of midrange contrast, however the tones in the highlights and the shadows are pushed toward the black and white. Consequently, highlight and shadow detail is obscured. Almost every amateur photograph suffers this to some degree, somewhat blown-out highlights and blocked-up shadows are practically hallmarks of the snapshot. The way to approach this is to fix a control point in the mid-point and then to create a reverse-S. Subtle adjustments work well, but stronger change may be needed if there is bad tonal separation in the highlights and shadows (very low Contrast). Strong changes can bring out good shadow and highlight detail, but will compress mid-tones and lower Contrast in that region, making the print look flat and lifeless. For example, mid-tone surfaces will lose Contrast as will skin tones (and faces can start to look flat). 

Ctein talks about a ‘slider’ feature called Shadow/Highlight (in Photoshop) which provides an alternative to using Curves. When using Shadow/Highlight he found that the highlights had more details, and the shadows were more open, with better colour and saturation than in the Curves-altered version, without destroying the mid-tones. Pixelmator has a number of ‘slider’ colour adjustments in addition to Levels and Curves, namely Brightness/Contrast, Exposure, and Lighten Shadows/Darken Highlights (which I guess could be the same as the Shadow/Highlight option Ctein mentions). Ctein felt that the Shadow/Highlight was effective because it selectively works on the highlight and shadow areas but leaves mid-tone areas alone. However the Lighten Shadows/Darken Highlights in Pixelmator is just 2 ‘sliders’ whereas the Shadow/Highlight shown by Ctein has 8 ‘sliders’, so considerably more flexibility (I guess it is the difference between a one-off 30€ and an annual subscription of at least 360€). 


Above we have a standard Curves plot, and below we have the same photograph but adjusted using the Lighten Shadows/Darken Highlights (20% on each slider). What the sliders appear to have done is to dramatically reduce the presence of shadows (blacks) and highlights (whites), whilst retaining mid-tones. 


Many faded prints have very poor Contrast in the shadows. Even when you restore the overall Contrast range of the print you may not have a good tonal separation in the darker areas. Control points can be placed on the line to do two things. They can be placed in the darker region and lifted to increase Contrast in the shadows and lighten them to bring out more detail. To restrict the lightening effect nearer the mid-tones and highlights control points can be added further along the curve to keep it from arcing upward overall. Locking down values near the highlights ensured that they didn’t lose any Contrast at all. It increases the Contrast a lot in the shadows and sacrifices a little Contrast in the other tones. It will also make the print appear somewhat lighter overall. 




As an example, what I have tried to do in the above photograph is to widen the tonal range by bringing black and white nearer together. Also I fixed the control point in the mid-tones. I tried to increase the Contrast and lighten the shadows to get some detail into the trousers. I also tried to darken the highlights to try to get more detail and shadow in the snow on the ski slope.


Below I have tried to do the same. My objective here was to try to see the tea in the cups, darken the jacket, and keep skin tones fresh.


 

And below what I have tried to do is to remove the excessive highlight on the face, and improve the Contrast so as to see some detail in the collar of the pullover. 



Many b&w photographs from the mid-20th C have both low Contrast and improperly exposure (and faded as well). Photoshop allows a sophisticated response to these problems, but what can we do with Pixelmator? 

Below I have used the Exposure slider to reduce the exposure by -25%. I have then modified the Curve by bringing the white and black points closer together (to extend the tonal range) and increasing the slope to improve the Contrast. I finished with a slight reverse-S to try to give the pullover more texture and to reduce the highlights on the face.


 

What we can see with these different approaches is that there is considerable freedom in the way we can manipulate Brightness/Contrast, Exposure, Lighten Shadows/Darken Highlights, and Curves. In each iteration I felt that the different photographs looked better, without really knowing what better meant, and without knowing what was the range of possible improvements. As Ctein often stresses, experimentation is key, and there is no perfect or best restoration, it is a matter of what is possible and what is the most pleasing to the ‘client'.


I think it might be quite useful to have a series of reference or target images, in order to show what a 'perfect' example might look like. 


In his book Ctein extends his discussion on Restoring Tone well beyond what I have covered above. I have decided to stop here until I have acquired more experience using Pixelmator on “real-life” photographs from my collection.  


Restoring Colour


A photograph with a rich tonal range from near-white to near-black will usually have good colour. When the tones are richer, the colours are richer to. When the colours are faded, the whites and blacks are not clean. Each of the RGB channels behaves like the grey channel in b&w photographs. 

The key to good colour is to adjust each of the RGB channels individually. And that means having broad well-populated histograms for each colour channel. The smart rules include the fact that most colour photographs have at least one fairly saturated colour. Each colour channels should have a broad set of tones spanning most of the range from black to white. Colour and tone are related, and one can’t be changed without changing the other. Rich colours are usually associated with dark values, pastels with light values. Contrast and colour intensity (saturation) are linked. Increasing Contrast of an RGB image will increase colour saturation. So get Contrast right before fixing saturation. Never boost saturation before getting Contrast right. Get the blacks, whites and middle greys right and the colours will follow. 

Ctein uses a reverse-S curve to make skin tones look natural, although usually insufficiently saturated.  


Correct tonality and contrast before fine-tuning colour.

 

Ctein develops in his book the restoration of colour using a sophisticated set of tools. Below I will use the Pixelmator tools, and we will see how far I can get. 



Above we have the original scanned photograph without any corrections, and below it the same photograph with each of the RGB colour channels expanded to improve tonal range, and also improve Contrast (i.e. steeper Curves).

Pixelmator offers a series of colour related sliders, hue (saturation and lightness), colour balance (cyan/red, magenta/green, yellow/blue each for shadows, mid-tones and highlights), as well as channels (red, blue, green and ‘constant’). I was not able to adjust these different slider options to create something even close to the photograph above after a “simple” RGB colour channel expansion. 

The below photograph involved first improving Contrast and widening the tonal range (improving the detail in the trousers and seeing more shadows in the snow), followed by expanding the individual RGB channels. 

Besides the blue tint to everything, it is difficult to assess the success of the modifications. As Ctein points out, colour is all about fine-tuning the aesthetics of a photograph. Certainly an improvement over the old, faded original, but is it the best that I can do? Experience will tell. 




Increase sharpness and fine detail


The ideal candidate for detail enhancement is a photograph that has good tonality and little visual noise, e.g. no heavy texturing, visible grain, or extensive scratches or cracks. Minimal noise is important because all these techniques enhance fine noise along with fine detail. The best of them work mostly on the photographic detail and don’t enhance noise too much, but there will always be some increase. If you start out with a visually noisy original, you will not have a good-looking photograph after you’ve sharpened it.

Equally there is a risk of 'over sharpening', and much can be achieved with shadow/highlight filters which make photographs clearer if not sharper and Contrast improvements with Curves that can reveal details much more clearly.


Adding sharpness and edges to the photographs shown on this page produces the following …





Certainly these photographs are not perfect, but what I like about each is that I know the people and I can 'see' them in the restored photographs.


Putting it all together


Here is my very first attempt to put all this together in one photograph. I started with a 300 dpi scan (TIFF) of a good quality photograph taken perhaps 25 years ago. 


Screen Shot 2017-09-08 at 09.16.03


Now I adjusted the Curves (RGB and individual colours) to try to increase the tonal range. 


I then increased very slightly the Brightness/Contrast sliders.


And finished by Sharpen (slight increase) and the application of Edges. Just below it is the original untouched image for comparison.  



Conclusion - The edited image is much fresher, brighter with better colours. Possible just a little ‘over-the-top' with colours a touch too intense, and the clouds in the sky have been lost. The shadows are also a little too deep. 


Below I have tried to be a bit more subtile in the editing. I wonder which one will print better? What do you think?





bernard.smith@mac.com  © Bernard Smith 2017-18