Dear Colleagues,
Texture analysis is a computationally very tedious process with myriads of output which may be too correlated for any further analysis. In analyzing texture using the MaZda software, the Fisher method is used to select texture features for further analysis such as classification. Fisher aims to reduce intraclass difference and maximize interclass difference. Does the method suggest the texture features selected best describe an image or a selected region of interest in the image?
In the best of my knowledge, the fisher method is used to performs multi-thresholding on image!
Fisher discriminant analysis or Linear discriminant analysis can allow you project an high dimensional space into a smaller space in which there is "hope" to separate the different classes more easily.
In that sense, it will not select feature dimension but project the data inside a new space by combining linearly the original feature based on the two criterion that your cited in your question.
In order to select the most significant feature, it is more appropriate to check the different methods for features selection (t-test, mRMR, etc.)
The Fisher's test is a statistical tool used to determine exactness of features. In my opinion, this implies that you may use the method to discriminate between two closely related features. You may consult an expert statistician, however.
Hi..
FDR is good method...
Feature Selection is basically selection of a subset of features from a larger pool of
available features. The goal is to select those that are rich in discriminatory information with respect to the classification problem at hand. This is a crucial step in the design of any classification system, as a poor choice of features drives the classifier to perform badly. Selecting highly informative features is an attempt
(a) to place classes in the feature space far apart from each other (large between-class distance) and
(b) to position the data points within each class close to each other (small within-class variance).
For more details: see chapter-4 of below book:
Book: An introduction to pattern recognition: A MATLAB approach
BY: Theodoridis, Sergios
You can try simple methods like Feature selection based on Fishers Discriminant Ratio, Class separability measures like Divergence, Scatter Matrices, Bhattacharya Divergence.
You can also try WEKA data mining tool for feature selection, as in WEKA different feature selection algorithms are there.
Gud luck...!!!!
- Badges
- Science topic