KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Settings

Author: search.filters.author.Puyati, WayoSubject: search.filters.subject.Principal component analysis

Search Results

Now showing 1 - 2 of 2
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Item,
    Efficiency improvement for unconstrained face recognition by weightening probability values of modular PCA and Wavelet PCA
    (2008-05-29)
    Puyati, Wayo
    ;
    Walairacht, Aranya
    Principal Component Analysis (PCA) is a well-known classical appearance-base method in face recognition. In the previous works, the preprocessing process significantly improved the recognition rate. Modular PCA and Wavelet PCA are the preprocessing processes of PCA, which increase the recognition rate of the original PCA. Modular PCA is suitable for the highvaried face database, while Wavelet PCA for the low-varied face database. In this paper, we propose the preprocessing method which combines between Modular PCA and Wavelet PCA with the weightening probability values. The experiments are compared among our propose method, Modular PCA, Wavelet PCA and original PCA with face database from Yale, ORL and UMIST. The experimental results show that the recognition rate of our method is higher compared to the other methods and also support variety of face database.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Item,
    PCA in wavelet domain for face recognition
    (2006-11-17)
    Puyati, Wayo
    ;
    Walairacht, Somsak
    ;
    Walairacht, Aranya
    In this paper, the preprocessing process aimed to reduce size of input image by using wavelet transform before transformed image is sent to the process of PCA for recognition. We used ORL Face Databases from AT&T Laboratories Cambridge in the experiments. The results show that the 4<sup>th</sup> Order Symlets level 2 and level 3 improve the accuracy rate of recognition when compare among Haar wavelets, the 4<sup>th</sup> Order Daubechies wavelets, and Biorthogonal wavelets (orthogonal 6.8). In the case of overall processing time for training, the length of filter of wavelet is directly effect the time consuming. Since LL subband of wavelet decomposition becomes the input for PCA, the memory usage can be greatly reduced.