Performance Analysis of Different Feature Extraction Algorithms

Feature extraction is the method of defining a set of features, which will most effectively represent the information that is important for analysis and concurrence. Gray level cooccurrence matrix (GLCM) is an important method to take out the texture features in medical image. Gray level co-occurrence matrix can simply take out the texture under single scale and single direction. However its description of the texture feature is not so detailed that is not good enough to extract the feature by this method. Nonsubsampled counter transformation (NSCT) is a kind of algorithm. This use the iterative nonsubsampled filter store to achieve a series of multi-scale, multi-direction and translation invariant frequency field sub-image. It is an move toward to texture extraction fall under the category of pixel base scheme. Quad tree decomposition is to be proposed to achieve more accuracy and it can highlight the details of image, more feature details can be extracted from the visual important objects that from the monotone area of the image. The image retrieval performance is highly improved as compared with the pixel based method.