Selection / Painting

Selection and Painting

How are portions of images selected?

There are several ways to select a specific part of an image. Once a selection has been made, any changes you make apply only to the selected area. This can be very useful when you want to change the contrast, brightness, hue, or other aspect of one object while leaving the rest of the image untouched. When a selection tool is used to create a selection, all of the pixels inside the region are selected.

What are the types of selection tools?

Paint and graphics programs often provide several selection tools. One class of selection tool is the shape select tool. The "shape" can be one of several varieties, including rectangular, elliptical, and user-defined polygon. Another tool, known as the lasso, is essentially a user defined polygon with sides of one pixel. This means that you just draw lines with it, release, and a selection is made. To fill in the (usually) missing side, a line is drawn from the starting point to the ending point upon release of the pointing device.

Magic wand selection

The last type of tool is the most interesting and useful. It is called various things, but usually is known as a "magic wand." The tool allows one to click on an object, and the entire object, curves and all, is selected. Wouldn't that be nice? Well, unfortunately, that is somewhat of a best-case scenario. The tool selects objects that are the same color or similar in color. So, you wouldn't have very good luck selecting a person with clashing neon clothes.

Is there a way around that limitation?

Yes, there is a way around it. Some programs, in addition to having a magic wand, allow the user to set a threshold level. A higher level means that more diverse colors are selected. This makes it useful for selecting pictures of objects that are partially in the shade. The other helpful workaround is that most programs allow you to hold down a key such as shift during the selection. This key adds your selection to whatever is already selected. Some programs also have a key which can be held down to remove a selection from the current selection.

How would you write a selection routine?

You could write an algorithm to select objects much more easily than you might think. Internally, there are two ways to represent the selected pixels. One way is to keep track of all pixels which are contained by your selection tool. The other way is to track the polygons themselves. Although the polygon method is more complicated, it requires much less data to be held inside your program. For example, a large square selected by the first method could well contain 100,000 pixels. The polygon containing them contains four. That's 25,000 times less data!!!

What about the magic wand tool?

The magic wand tool would be the most fun (i.e. the most difficult) to program. Central to the algorithm is the code that determines which colors are "close" to others. A good way to accomplish this is to give points for each color that has close to the same value as the predominant color of the selected pixel. If you select a red pixel, therefore, give points to colors which have similar red values. Then, out of these pixels, subtract points for differences in the other colors (green and blue in this example). Any points that are below a certain threshold are excluded.

The other necessary step is the order in which you select pixels. The best way is to start from the selected pixel, apply the algorithm to the left until you hit the first pixel that doesn't match your threshold. Use this value as your starting pixel, and do the same procedure moving towards the right. Then, move the search to the next line down or up (or both). If you can create threads to do the searching, "both" can be easier. Otherwise, you need to mark the line you started on and resume it when you get to the top.

Return to main page.