Key portions of the image assembly work done to produce the photo-art here below and here (but later) rely upon the blending of individual image layers using Photoshop. In essence, two or more different versions of the starting image are superimposed so as to gain an output which contains some of the features of each separate component. Such superimpositions have been possible since early editions of the Photoshop software program. Several of the modes available for blending are, however, newer to Photoshop and its users than are others. In tutorial material they are most often (all) illustrated by using two entirely different images (to demonstrate the superimpositions); fully different by virtue of content and coloration, on more-or-less a random basis. Given the complexities and subtleties involved in the different (sometimes even related …) blending operations, this can be the source of a great deal of confusion, even nightmares, at the beginning (never minding the end ..!).

This post just includes an outline description of a far simpler set of guidelines for analyzing the blend mode operations. Subsequent posts here – all tagged under the Photoshop category – will examine each successive mode, one at a time.  Additionally, I’ll try to capture some of the “fine tuning” options available in specific cases. Hopefully this will all be made simpler via the use of a set of related (actually correlated) images.

It should be noted, as just a part of this introduction, that all of the blend modes available in the newest version of Photoshop (CS4) rely upon strict computational mathematics. But, in addition, that this mathematics – which is proprietary – makes use of color models which are not used regularly directly at the user interface. For example, the majority of the Photoshop work I do is monitored, in output form, via the standard RGB color model, with all three channels active. I suspect this is a workflow option selected by the majority of users. But, the “behind-the-scenes” computations that Photoshop makes when blending operations are invoked may rely upon individual channel information within an (also proprietary?) HSY model (see, for reference, here).

Therefore, a tutorial which attempts instruction via the mathematics involved will run into a different complication. And, potentially, much bafflement. So, in a search for simplicity (at least to start with here), I’ll be using the image layers shown in the following two screenshots and, initially, in the black-and-white-only versions included there – without any accompanying algebra.

Here the layer designated as the overlay is simply a flipped version of  the layer underneath it in the stack (the underlay).

Somewhere below and earlier (i.e., here) there’s a slideshow made up of individually digitally filtered images, all derived from a single, starting image. Back in 2005 I wrote a short online article (it’s here: http://www.dpandi.com/howtos/artmasterpro/) covering some aspects of such filtering. Unfortunately, the source of the Photoshop-independent image processing software featured there – the Fo2PiX company – is no longer in business. But, nevertheless, a number of other filtering tools have become available in the intervening period, up to and including full “auto-painting” routines.

A key individual function of the Fo2PiX software was to introduce pictorial simplification. Basically this was achieved by the elimination of fine structural detail in the starting image (and the associated averaging of color across the larger, more prominent, contiguous areas of the image which remained). This was targeted – as an artistic objective – in order to emulate what traditional painters do when adopting an impressionistic style.

Nowadays much the same functionality is available in the Topaz Labs Simplify plug-in. One neat feature of this newer software product is that the degree of simplification produced has a set of fine sensitivity settings which scale correctly with the overall size of the starting image. However, one series of settings will provide a rendered image with a fixed degree of uniform simplification for the image as a whole. There is no ability to directly restrict the amount of simplification in one or more areas of the original other than by creating a mask within Photoshop (usually before launching the plug-in).

One consequence of this which occurs with images where the subject has been deliberately isolated (or “cut-out”) from its original background is that some of the simplifications at the edges may remove and/or distort the isolation.  The first two images below have been cropped from an isolated version of a starting image introduced earlier and re-worked using Simplify. It is easy to see there that portions of the (white) background have now become blocks of color derived from the floral portions of the image. By using the mask – the one which was first built to isolate the full bouquet – we can arrive at the last re-rendered version of the image included here (the third in the group below). Now the “block” simplification of the floral components at the core of the subject material is independent from the definition of the edge. This is different from, and, perhaps, better than, what a conventional painter might choose to do in portraying the same type of subject (i.e., a floral still life). More will be written later on forcing distinctions between the edge and core regions of digitally re-rendered images.

some baseline and uniform simplification

an increased amount of uniform simplification

the same  increased amount of uniform simplification, but using an isolation mask – just two steps from the original photograph