Feature extraction of X-ray chest image based on KPCA

In view of the nonlinear image information loss and lack of characteristics which is caused by principal component analysis in the feature extraction process, an X-ray chest image feature extraction method based on KPCA is proposed. Original feature space is mapped by kernel function to a new space where dimension reduction is implemented and features are extracted, and then nonlinear information is converted to linear information in the feature space. This method reduces feature dimension considerably while it maintains adequate original X-ray chest image information. Experimental results show that this method can enhance retrieval accuracy and has better performance than principal component analysis.