Add a pre-seg procedure to help to decide better parameters of segmentation

[LLeiSong]

Now the package only allow the user to set the parameters of segmentation. Without any pre-knowledge, the user might need to tune back and forth to get the best segmentation result. Next step of this package is to add a pre-segmenting procedure to analyze the optimized parameters of segmentation. So:

• If the user only has a few images to segment. Then the user could manually set the parameters to get the best result.
• If the user has many images to segment that are impossible to manually check. And the objective of segmentation is not to get the optimized segments, but to use the segments to do other analysis (e.g. clean the classification results). Then using pre-seg procedure to pick the parameters is a better option.
• Or if the user has ground truth labels, then the user is better to analyze the relationship between ground truth and segments to get the optimized parameters of segmentation.

[LLeiSong]

So this would be similar to unsupervised classification?

[LLeiSong]

After some experiments, the characteristics of each objects do not have any consistency with the size of objects. Keep it open for more details.