Summary: | 碩士 === 義守大學 === 資訊工程學系碩士班 === 97 === Principal component analysis (PCA) is a linear transformation based on linear algebra technology using less data to explain the original data with least errors. It is usually used in signal processing to reduce the dimension of information, Because PCA algorithm requires the computation of eigenvalues and eigenvectors, it is time consuming. In this paper, we use artificial neural networks (ANN) to estimate the principle eigenvectors by learning characteristics. The traditional vector quantization (VQ) uses a codebook to represent all possible image blocks. The advantage of VQ is the high compression ratio and the shortcoming is the codebook design which is time-consuming, In PCA, the eigenvectors, also called the eigenimages, also forms a codebook in another sense, which is similar to VQ scheme. This paper uses the eigenimages as a codebook and performs a VQ-like codng method to take the advantages of both PCA and VQ methods.
|