Dimensionality Reduction of Image Feature Based on Mean Principal Component Analysis

Abstract

With the development of computer vision and various advanced medical imaging devices, medical images contain more and more information. It is difficult to accurately express the characteristics of medical images with a single image feature, and it is not convenient for later diagnosis and recognition processing. In this paper, the chest CT images of children were used as research objects. Its colour features, texture features and shape characteristics were extracted. The main component analysis method is an effective method to reduce the characteristic redundancy. However, considering the limitations of principal component analysis in data standardization, in this research, the method of mean principal component analysis is used to reduce the characteristics of extraction. Experiment result shows that this method achieves a higher contribution rate with fewer dimensions. The conclusions of this research are beneficial to image recognition research.