Content-based image retrieval for not-well-framed images using multiresolutional eigen-features

A content-based image retrieval system is implemented using principal component analysis (PCA) and multivariate discriminate analysis (MDA) with several different well known feature vectors, such as Radon, Gabor, and wavelet representation including raw and simple modified histogram feature for the purpose of the comparison. The image data set used in this project is characterized as "not-well-framed" which means that there is no limitation of variation in the size, position, and orientation of the objects in the images. The performance of methods (PCA vs. MDA) and each feature vector are compared to each other.