In addition to specifying a mask for a pattern training image, as described in the section Pattern Masking, you can specify a second mask for the run-time image. Any mask image you specify must have the same dimensions and image offset as the run-time image.
PatMax will exclude from the run-time image any features that lie within don't care pixels. Don't care pixels are defined as all pixels with values less than 128. Pixels with values greater than or equal to 128 are considered care pixels. Features that lie within care pixels are retained.
Features that are present in the trained pattern but masked from the run-time image will reduce the coverage score received by found pattern instances, as shown in the following figure:
Features in the run-time image that correspond to don't care but score portions of the trained pattern will be scored as clutter features, as shown in the following figure. This is because those features are not part of the trained pattern.
You can prevent PatMax from treating those features as clutter features by using the don't care and don't score mask value when you train the pattern. For more information, see the topic Pattern Masking.