Segmentation is a complex task, and the number of functors in the LTI-lib that deal with segmentation grows.
.. A few extra-notes are required.
There are two sorts of segmentation functors. Traditional "complete" algorithms and "blocks", with which you can build your own segmentation approach. There are also tools to convert the results from one representation to another.
The complete algorithms are:
- Threshold segmentation. Used to separate background from object (two-classes segmentation). You give a "window" for pixel values that should be recognized as object. The rest will be background. Some algorithms can choose automatically the threshold values. (see lti::thresholdSegmentation, lti::optimalThresholding)
- Watershed segmentation. This sort of algorithms are widely used. Several implementations can be chosen using the functor parameters (see lti::watershedSegmentation)
- Comaniciu's Mean-Shift algorithm is also implemented. With it you can choose between quantization, oversegmentation and undersegmentation. (see lti::meanShiftSegmentation)
- Region growing for a two-classes segmentation problem can also be found here (see lti::regionGrowing)
There are several blocks, that can be used as part of a more complex segmentation approach.
- A statistical background model per pixel can detect a continuous background from a moving object or objects. (see lti::backgroundModel)
- A pre-segmentation approach is implemented, which finds regions using color quantization and marks as background the colors most used at the border in lti::csPresegmentation.
- Similar to the pre-segmentation is the lti::kMeansSegmentation, but here no background is identified, only the color quantization and (optionally) the smoothing stages are done.
- lti::whiteningSegmentation uses the kMeansSegmentation to segment the original image and a transformed one, and combines the results of both.
- Given a list of area points, where each object (or region) is represented by an lti::areaPoints object, it is possible to generate a so called similarity matrix using the lti::similarityMatrix functor. Here a similarity measure will be generated for all adjacent objects.
- With a similarity matrix it is possible to merge similar regions using the lti::regionMerge functor.
Segmentation algorithms produce usually so called "masks", i.e. channels, where each pixel contains a label value. These masks can be considered for two classes problems, where only values of zero and not-zero will be differentiated, and "labeled" masks, where each value is considered as a different object. To extract the position of each single object in the image in an efficient way the functor lti::objectsFromMask should be used. Sometimes the lti::fastRelabeling functor provides all functionality required for this task.
The lti::boundingBox functor can use one of the pointLists extracted with lti::objectsFromMask to produce a new image with only one object.
It is useful sometimes to generate a color image, where each label found in the segmentation is replaced by its mean color, producing an color quantized image. Here, the functors lti::usePalette and lti::computePalette are helpful.
