Published May 1. 2002

Digitizing flies, editing

By Martin Joergensen (supported by Steve Schweitzer)

Handling you electronic pictures properly and using the right methods and tools can greatly enhance the results.
This fly was scanned directly in the flatbed scanner. The top picture is unmodified and not compressed while the bottom one has been sharpened, heavily compressed and loaded and saved again a few times. The unmodified pictures especially excels in tonal richness and subtlety and even though the bottom one might seem sharper it does have less details.

Working with originals
Two worlds
Before commencing we want to stress one particular fact. Please notice that there is a huge difference between the demands on pictures for web use and pictures for work and storage. And these demands work against each other. The practical meaning of this is that you have to have two sets of pictures: one for work and storage and one for publishing on the web.
All work should be done on a working set of pictures, which are stored in an optimum quality. The last step before publishing is converting to s suitable publishing format like JPEG or GIF. Do not work on JPEG's as they deteriorate every time you load and save them.

Your most important asset when working with digital photos is your original. Never modify the original! It is like burning your negatives. This particularly goes for photos taken with digital cameras. If you have slides, prints or negatives, you can always scan your pictures again. You can also reload a Photo CD, but you can not recreate a digital photo once it is modified. Store your originals in separate folders, burn a CD, write protect them - just don't work on them. Load them into your photo editing program and use Save As immediately or make copies of all relevant pictures and work on the copies.

We recommend working on an oversize picture. Aim for somewhere between 1.5 and 2 times the final size. If your goal is a 300 by 200 pixel JPEG, work on something like a 640 by 480 pixel original. Less can do, but size will give you quality.
The reasons for this are somewhat obscure, we admit, but in the process of optimizing the picture for the web, we often experience that it is easier to get a crisp, detailed photo when scaling down as opposed to scanning in the final size.
Really large pictures, like the high grade Photo CD formats are excellent for many purposes, but choose one of the lower grades if your final destination is the web exclusively. This will save you time in the process unless you're working on a 256 meg G4 or and turbo loaded 500 MHz Pentium III.

Simply cropping the picture can enhance the subject. Here the fly has been rotated and cropped close before scaling to its final size.

Manipulating pictures
You will generally want to use the RGB color format for your working pictures. This is the generic screen format, and no matter what the end result, this is the optimum format for screen work. Convert CMYK and palette pictures to RGB before going on. This will ensure a proper presentation on your screen, and an optimum result when using filters and effects.

Crop your picture before working on it. Removing surplus will save file space and work time. Cropping is usually done with a special crop tool, but most programs support cropping to the edges of a selection.
Use the cropping to tighten the composition of your photo. Go as close as you can, still preserving the important parts, but enhancing the composition of the picture. One of the most common errors in photo work is not cropping close enough.
Some programs like Photoshop will preserve the entire picture, even after cropping. Don't worry. Photoshop knows what it is doing, but if you insist on worrying anyway, mark the part you want to keep, copy it, and paste it into a new picture.

By using a level or histogram function it is possible to remove shadows and gray tones from the background of a fly scanned on a white background. Try passing your mouse cursor over the picture to swap between the original scan and the edited copy.

Tonal range
Few pictures are perfectly exposed from the beginning, even fewer are perfectly printed and you very rarely see perfect scans. All this ads up to pictures, which have an inferior dispersion of light. This can mean several things, the most common being poor contrast - too much or too little.
The second most common problem is color errors. Many pictures are off balance, being too red, green or whatever.
The third most common problem is too light or dark pictures lacking tones in a whole region rendering them quite flat to look at.
None of these errors are easy to adjust if they are very pronounced, but if they are within reason they can be all be removed.

If you really want to understand picture adjustment, you want to understand tonal histograms or tonal ranges. A histogram graphs the number of pixels at each brightness level in an image. It is shown as a bar diagram with each vertical line representing the number of pixels at a certain level of light. The well-exposed picture has an evenly distributed range of pixels in almost all brightness levels. The distribution should be without peaks - especially in the extreme darks and lights. By analyzing the tonal range of a picture, you can pinpoint its faults and choose the right adjustment tools. We have included a few examples of pictures with certain errors and their respective histograms.
The most powerful tool in adjustment of histograms is a tool such as the Levels tool in Adobe Photoshop. This tool will let you manipulate the tonal range of a picture by cutting off the extremes and redistributing the existing tones within the whole range - totally black to absolutely white.
The level tools are particularly handy if you want to isolate objects on a totally white background. By stretching the lighter grays you can turn them into a completely white. In this way the light shadows around the object will disappear.


The normal picture has a normal RGB histogram (above) and evenly distributed R, G, and B histograms (right)

The red picture has a normal RGB (above) and R histogram, but few G and B tones (right).

The blueish picture has a normal RGB histogram (above), but low level of R tones (right).

Color balance
Pictures should contain a normalized range of colors and not have any tendency to slide towards a certain color. In extreme cases pictures with color errors can seem tinted, but usually they will just seem warm or cold.
Good examples of common color errors are underexposed winter pictures, which come out light blue, or indoor pictures in either bulb or neon light, which typically come out yellowish warm and greenish. The latter will show faces in a particularly grim hue. Sometimes such errors can result in interesting effects, but usually you want to remove a bit of the tint, and give people their natural skin color.
Again the histogram will reveal the details of such errors. Check each channel separately - Red, Green, and Blue (R, G, and B) and make sure that all curves are as close to each other in tonal distribution as possible.
Many photo editing programs have automatic tools for these adjustments, but using a tonal curve tool and a levels tool to manipulate each of the R, G and B channels of your picture will give you better control and ultimately a better result.

Too low contrast will yield narrow distribution of tones. The above picture has light areas, but no dark ones. Stretching the range will give better tonal distribution, but no further details.
Too high contrast will result in a saddle shaped histogram. Nothing will cure this as we lack the details in dark and light areas of the picture.
The normal picture will have a normally distributed and rich range of tones. This is a bit on the dark side.
Contrast and brightness
The plain contrast and brightness controls that even the most primitive photo-editing program contains are crude and simple tools compared to the above mentioned facilities. They can be used for some quick adjustments of the most obvious errors, but will usually bring you too far within very narrow ranges of adjustment.
Too little contrast can be corrected with fair results, while too much contrast usually means lack of details in dark areas and highlights. These details can not be reconstructed with automatic functions.

Correcting errors
You now have a picture, which has the best possible tonal range. This forms a good foundation for the following work, which encompasses editing, filtering and manipulating the contents of the picture - the motif itself.
This is the stage where you correct errors such as dust, scratches and other physical flaws generated in the scanning and physical handling of the picture. Scratches and dust are the most common errors, but stains and curls are also seen.
The method is fairly similar for all of the above.
Enlarge the relevant section of the picture, so that you can see each pixel and use a Stamp or Brush function to paint over the spots or cracks. Do not use Smudge or Swab tools as they seem to rub out colors and leave clear marks. The idea is to mimic or copy the surroundings as closely as possible to mask the faults.
Pick up colors from the neighboring areas and paint over the spots. Frequently select a new, but close, color and use that in order to copy the blend of the adjacent pixels.
For larger flaws you can copy similar areas using feathered edges and paste the copy over the fault. By manipulating the copied patch, turning it, making it partly transparent, you can mask even large errors.

The various photo-editing programs have a large selection of filters for many different purposes. Most filters are effect filters, which will alter the photo - sometimes beyond recognition. But several filters are made for correction purposes and can help you in the process of enhancing the picture.
A few advanced programs have special filters for removing dust and scratches. More common are filters, which remove iterative patterns like printing raster or scanning lines from poor scans, and video or television pictures.
Blurring filters can sometimes be a good tool to use for removing errors. Use the blur filters with caution as they do exactly what the name implies - makes the picture less sharp.

Enhancing edges
Of all filters the sharpening filters are the most important when working with web files. As the resolution of the pictures is usually low and the size is limited, we can not have the level of details that is seen in prints or slides. The sharpening process will visually enhance the details in a way, which actually degrades the picture, but makes it look sharper - mainly by enhancing contrasts and edges.
Strangely enough a tool known as an Unsharp mask is also widely used for sharpening purposes. This filter will add a light and a dark line along edges and can give some significant enhancements in most pictures for online use.
The sharpening and unsharpening filters usually have several parameters, which can be adjusted before running the filter. Experiment with these parameters to obtain the best results. Use the filter, undo, adjust parameters and run the filter again. Repeat the doing and undoing several times to find the best setting. The best programs have live preview boxes, which can be used to judge the result before letting the filter loose on the whole picture.

A warning
Sharpening should be done with great care - especially before you scale the picture. As mentioned sharpening will deteriorate the picture, and this loss of quality can very well be enhanced in the scaling process.
Log in or register to post comments