One of the widely used approaches in image recognition is principal component analysis (PCA) because of the good balance between the simplicity and speed of the algorithm and the results obtained by using it. In the last years many variants of PCA were developed: two dimensional PCA, two directional two dimensional PCA, extended two dimensional PCA and extended two dimensional two directional PCA, the last one developed by the first two authors of the present paper. In this paper we go further with this study by considering a mixed approach between E2DPCA and diagonal PCA. The mixed approach not only takes approximately the same amount of time for training and testing as the classical approach, but also gives better recognition accuracy for some of the PCA algorithm variants.
展开▼
