| Customize Help

Steps to performing a pattern search


The following steps provide a basic methodology for using the MIL Pattern Matching module:

  1. Allocate a Pattern Matching context, using MpatAlloc().

  2. Load or grab an image from which to define a model. This image is known as the model's source image. This must be an 8-bit unsigned grayscale image that should be (or contain) a good example of a typical model.

  3. Define and add your model to this Pattern Matching context, using MpatDefine() with the model's source image.

  4. Optionally, repeat the two previous steps as many times as necessary to grab, define, and add your models to this Pattern Matching context. All models in the same context must be the same size and have the same resolution, but models can come from different model's source images.

  5. Optionally, mask any irrelevant areas of the model using MpatMask().

  6. Optionally, specify your required search settings for both the context and the individual model(s), using successive calls to MpatControl().

    When searching for several models, the search settings can either be set for the first model (when M_SEARCH_MODE is set to M_FIND_BEST_MODELS) or in each model (when M_SEARCH_MODE is set to M_FIND_ALL_MODELS). Note that, with M_FIND_BEST_MODELS, all models must be of the same size, and with M_FIND_ALL_MODELS, all models must use the same search region.

  7. Preprocess the Pattern Matching context, using MpatPreprocess().

  8. Allocate a result buffer to store the results of your search, using MpatAllocResult().

  9. Grab a target image. Optionally, process it to improve its quality.

  10. Search the target image for occurrences of the models in your Pattern Matching context, using MpatFind().

  11. Retrieve the required results from the result buffer, using MpatGetResult(). You can use MpatDraw() to draw the occurrences' bounding box and/or a cross at the occurrences' reference position.

    The coordinates resulting from a search return the reference position of the model, relative to the center of the top-left pixel of the target image. To determine what were the equivalent coordinates of the model's reference position in the model's source image, use MpatInquire() with M_ORIGINAL_X and M_ORIGINAL_Y.

  12. Optionally, save your Pattern Matching context, using MpatSave() or MpatStream().

  13. Free all your allocated Pattern Matching objects, using MpatFree().

In general, the first seven steps are performed once, while steps 8 through 11 are repeated as required. Note, in practice, models are usually saved on disk, using MpatSave(); therefore steps 1 through 6 are often replaced by a single step that restores a saved model from disk, using MpatRestore().