Days
Hours
Minutes
Mix of organic and geometric shapes in skies

Masking Techniques and AI

Introduction

I’ve written various articles on masking on this website. Very recently I wrote about holistic masking and how a good and efficient masking approach depends heavily on the type of adjustments you want to make in your photo. So first comes image analysis and the visual style of the image with its corresponding processing approach, and then comes the masking approach. Every aspect of image creation is part of a whole and interconnected and the entire process can therefore be called holistic.

 

Masking techniques weren’t addressed in that recent post, but they have been addressed extensively in the past. The techniques, however, have evolved over the years and AI technology has also changed the use of traditional masking techniques. Therefore, it’s time to list the most important masking techniques and what role they play in the masking process, now Adobe introduced AI technology to enhance this masking process.

Mask types

Before going into the masking techniques a few categorizations need to be made since the type of mask determines the technique to be used. 

Hard and Soft masks

The first type of categorization is the distinction between hard and soft masks (also called luminosity masks).

  • Hard masks: these are masks with a hard edge. There is no blending between a selected object or plane and its adjacent area and smooth blending is actually what is not desired. Therefore, these masks are needed to isolate and control shapes or well-defined planes. These masks consist of pure black, that which is not selected, or pure white, that which is selected.
  • Soft or luminosity masks: these are masks with no defined hard edge. The frame of the image is the edge. There are only so-called soft transitions. These masks are rarely pure white (selected) or pure black (not selected) but usually have shades of gray. These masks are needed to only isolate light, not shapes. However sometimes the light/shadow will coincide with the shape, but that’s a rare exception if your images are correctly exposed.

Geometric and Organic masks

The second categorization distinguishes masks for geometric shapes or planes from masks for organic shapes and planes.

  • Geometric planes or shapes: hard masks for geometric shapes or planes can usually be found in man-made objects like in architecture and demand a high degree of accuracy. The more you push the contrast the more accurate those hard masks should be
  • Organic shapes or planes: hard masks for organic shapes or planes are everywhere in nature and are very rarely clearly separated from the rest of the adjacent environment (ground). The masks need less or more accuracy depending on the degree of clear separation. The clearer the separation between the object to be masked and the adjacent area, the more accurate the mask needs to be. Rule of thumb: rough selections will suffice unless they intersect/overlap with the sky or any other ground or with a plane that is part of a figure.
  • Skies: skies can sometimes be geometric planes, e.g. in clean seascapes, or can be a mix of geometric lines and organic lines. In any case, sky masks need to have the same accuracy as masks for geometric shapes and planes, especially if you push the contrast harder.

Masking techniques

Before describing the masking techniques there’s a rule of thumb that you always need to remember before even starting to mask. And it’s not just a rule of thumb, it is also common sense, let’s say it is a rule of a giant’s thumb:

The more control you need over an object, a shape, or a plane, the more it needs to be isolated from the adjacent areas, and the more accurate the hard mask needs to be. If you don’t need control, do not mask, if you only need a bit of control, then mask it accordingly with a lower degree of accuracy or with much smoother (feathered) edges. 

Pen tool: this is the basic technique everyone should be able to use. The pen tool can be used for hard masks for geometric planes or shapes. It is often senseless to use the pen tool for highly intricate organic shapes or objects. Unless you require marginal control only.

Contrast-based techniques: contrast-based techniques depend heavily on the unadjusted RAW luminance information and the ideal exposure of an object. Channel masking is a popular contrast-based technique that is great for both geometric shapes and planes and organic shapes and planes. It’s a technique, just like the pen tool technique, that is very common and widely used, and instructional tutorials on these methods are widely found on the Internet. 

A better and far more accurate and powerful contrast-based technique is SMC (sectional multi-channel masking) that uses not only the RGB channels as in channel masking but also an array of luminosity masks in overlapping sections. It’s a technique that I developed in 2015 and is explained in this 5-hour video. As far as I know, no one else is teaching this technique. More than simple contrast-based techniques such as the channel masking technique, this technique is ideal for masking intricate details like hair and increasing accuracy. It also utilizes techniques such as mask-in-mask techniques for very sophisticated fine-tuning of masks.

Refer to the image below for a side-by-side comparison of the two contrast-based techniques.

Color-based techniques: as the name implies this technique depends on color information only. This technique is only useful if an object, plane, or shape contains enough unique and uniform colors. The more the color varies the less useful and accurate it is. Having said this, if you’re dealing with blue skies, then this technique will do very well to mask skies. Mainly I use this technique as an additional technique if all other techniques fail or aren’t sufficiently covering the entire mask.

Generally speaking, contrast-based techniques are better, more accurate, and more versatile than any other technique for hard masking. The table below will clarify that statement.

Adobe’s AI technology

With the introduction a few years ago of AI technology in Adobe products to make masking easier one might question if it’s still needed to know any good masking techniques when Adobe’s selection features like Sky selection, object, and subject selection will do the job for you. 

The short answer is ‘yes’ because Adobe’s AI-supported selection features are still not perfect yet. And if and when they will eventually be perfected, they are always targeted at average and more mainstream adjustments that don’t require the accurate hard separation of high contrast B&W processing styles where contrasts can be pushed very hard. 

Just take a look at the sky selection feature: they always have soft gradational edges around the horizon. This may work well enough for average adjustments in images that have only organic shapes, but they become a problem where man-made objects are to be processed in high contrast and 3-dimensional visual styles.

So, manual correction is still needed, even when the AI-supported selection features have been perfected by Adobe. Therefore, knowing all the manual non-AI techniques for creating masks will always be valuable for customizing the masks to one’s individual processing style. But obviously, Adobe’s integration of AI for masking will save you a lot of time. You just need to fine-tune and customize it to your liking now.

Quick Mask Pro plugin (Artisan Max)

The plugins for masking I developed, Quick Mask Pro (recently also Artisan Max), and are based on the sectional multi-channel method, were originally designed to create masks from scratch, pre-AI, but now have been re-designed to fine-tune and customize the AI-generated masks and create variations with partially smooth and partially hard edges to accommodate for a variety of situations with a very high degree of accuracy. To get an idea I’m referring to the many videos I created on the plugin on my YouTube channel

More info: Video or Online Class

How to know what technique to use when and to what extent can masks further be fine-tuned and customized? This is a question that can only be answered satisfactorily with visual demonstrations. It is too visual to describe in words in a post. You can learn from me the techniques on the self-developed technique of sectional multi-channel masking via a 5-hr video tutorial or the same but now also how to combine it effectively with Adobe’s AI technology, live via the online and interactive masking class I just announced for April 28, 2024. Click the buttons below.

Share this post

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.