Topic: Digital image processing – standard procedures

The following contribution describes the standard procedure for, for instance, digital image processing reproduction photography, which serves as a good basis for the purpose in hand:

1. Capturing the image

When scanning an image, it is not possible to improve quality by scanning using a high resolution and then reducing the image subsequently to the required size. This doesn’t apply when the reduction method used is better than the scanning function – it is not possible to say in advance which approach will produce the best results, as this depends on the combination of the scanner used, the scanning software and the image processing software. A reduced digital image must always be sharpened retrospectively. It is always better to use RGB instead of CMYK as it represents a more comprehensive color space, unless there are good reasons for not doing so.

With digital cameras the situation is a little different: The image reduction techniques which are employed in a digital camera’s internal processing are less powerful than those which can run on a modern PC. Therefore the best results can be obtained by selecting the highest possible resolution while photographing and then reducing the images on the PC.

2. Clipping (or isolating objects)

Image edges which will be removed later falsify histograms, i.e. they give rise to tone values – brightness values – which are not present in the actual image. They also increase file sizes which slows down processing, especially if the computer has too little main memory. When scanning, try to set the prescan frame as accurately as possible.

3. Tone value correction ('brightness range correction')

Fine adjustment begins with spreading out the tone value range. This function is available in PhotoPerfect’s contrast functions under 'Automatic contrast HSV'. This is followed by the gamma correction, in order to set contrast and brightness values and to correct underexposure or overexposure. Only then should one start to deal with overdark shadows (dark image sections) or too bright highlights (light sections) using the appropriate tools, e.g. ‘Dimming/Brightening’, ‘Brighten shadows’ (which can also darken highlights) or ‘Selective dimming/brightening’ (which can also be used to correct highlights..

4. Color correction

As the brightness correction also leads to a color shift, the color correction is usually performed after the brightness correction (see 3.). In color correction or color balancing color casts are removed – this includes white balance and color temperature. The causes of the errors concerned are wrong lighting or camera settings. It is also possible to intensify or soften colors in selected image portions – in PhotoPerfect with the color curves, for instance. The latter is referred to as gradation. The tools can be used separately for each primary color or in any combination of them.

If all the primary colors are altered simultaneously, then this is equivalent to a straight tone value correction as described in section 3. That is to say, in that case brightness, contrast and mid-tones (mid range brightness values) are set using a gradation curve. As required a noise suppression filter can be used as well. For the subsequent processing described in section 5. it may be advantageous to perform some softening.

5. Mounting and retouching

This includes enhancing the image with a suitable background, removing unwanted objects from the image and correcting red eye effects, hiding mistakes and so on. With very elaborate operations of this kind, one may reach the limits of PhotoPerfect’s capabilities. Almost everything is possible, but sometimes only at the cost of a great deal of time.

6. Sharpening

The image should be sharpened immediately before printing, output to other media or digital presentation in web or slide shows, as sharpened images can practically not be corrected or retouched. Sharpening increases contrast. Transfer to a print medium leads to a renewed degree of softening, so for these purposes images need to be oversharpened. The question as to how much depends on the output medium – e.g. more for rough paper than for smooth.

PhotoPerfect implements an extremely powerful unsharp masking filter in a series of partly automated steps. The filter creates an unsharp copy of the image and overlays it with the original. The contrasts at the boundaries between neighboring colors are enhanced, while saturation and color gradations remain as they are. In addition, there are other processing techniques which are only available in PhotoPerfect and which sometimes produce even better results. Further details are available here: Correct sharpening And Xe847 / I2E?

These automated routines are mainly relevant for section 3 and most of 4.

The topic image output is too large to be dealt with here. It involves items such as printing, dpi resolution, image services etc. (There is much information available on these subjects and we plan to provide articles in this forum in due course. - Editor)

The Extension: Digital image processing – step by step order of execution

Another user recommends performing pincushion correction if necessary BEFORE isolating or clipping when an ‘automatic correction’ is desired, otherwise the image will become overcorrected.

Tip from author J.K. from another part of the forum:

He sharpens before executing functions which cause a loss in sharpness:

e.g. with pincushion correction (!), rotation other than through 90°/180° or size reduction.

Consequently, these alterations should be performed as step 7 after the steps detailed above. This could mean, however, that the pincushion correction collides with the edge clipping: Either the pincushion correction is then too much if a lot of the image is clipped away first, or the image takes on a degree of unsharpness if the pincushion correction is carried out first as recommended above.

An exact solution to this dilemma would only be possible if PP could process the image at step 7 as if it were still as large as it originally was. That would mean that the pincushion correction would be performed as if the phantom edge were still there, so that the inner part of the image is corrected with the right degree of intensity.

Suggestion from another user: This pre-sharpening process can be helped by implementing macros, e.g.:
- sharpen a little, then automatic pincushion correction
- sharpen a little more, then reduce to eMail format


When sharpening, the screen display for images should always be with 100% zoom factor. Images are usually displayed on screen with 72 ppi, but are quite likely to be printed at 300 ppi or more. On output, several data file pixels are used to compute one matrix dot so as to avoid pixel structures. However, this leads to a generally softer overall impression. So on screen images which are to be printed should look somewhat oversharp. The same thing applies, the worse the quality of the paper.

Here some rules of thumb for USM:

Small radius, high intensity for fine, sharp subjects. With unsharp ones, however, it just increases blemishes and skin tones become patchy.

Large radius, high intensity for unsharp images.

Large radius, low intensity as a good starting point for extreme magnifications.

Average settings for radius and intensity are ideal for skin tones.

According to Wargalla, radius settings above three pixels in size spoil any normal image. It is incomprehensible why PhotoShop, for instance, allows up to 250 pixels. Edges get emphasized too much and there is too much loss of detail.

A high threshold value (see link above) also requires a higher degree of intensity. If both are too high, areas are no longer sharpened and edges may look unnatural. What is required is a harmonious interplay of the factors threshold, intensity and radius – thus by using the threshold value it is possible to prevent areas from being sharpened.

Cheers

Ralf Eberle