The nonredundant contourlet transform (NRCT): a multiresolution and multidirection image representation with perfect reconstruction property

Zhang, Shenqiu (2008) The nonredundant contourlet transform (NRCT): a multiresolution and multidirection image representation with perfect reconstruction property. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (10MB)

Abstract

Multiresolution and multidirection image representation has recently been an attractive research area, in which multiresolution corresponds to varying scale of structure in images, while multidirection deals with the oriented nature of image structure. Numerous new systems, such as the contourlet transform, have been developed. The contourlet transform has the benefit of efficiently capturing the oriented geometrical structures of images; however, it has the drawback of a 4/3 redundancy in its oversampling ratio. In order to eliminate the redundancy, this thesis proposes a progressive version of the contourlet transform which can be calculated with critical sampling. The new proposed image representation is called the nonredundant contourlet transform (NRCT), which is constructed with an efficient framework of filter banks. In addition to critical sampling, the proposed NRCT possesses many valuable properties including perfect reconstruction, sparse expression, multiresolution, and multidirection. Numerical experiments demonstrate that the novel NRCT has better peak signal-to-noise performance than the traditional contourlet transform. Moreover, for low ratios of retained coefficients, the NRCT outperforms the wavelet transform which is a standard method for the critically sampled representation of images. -- After examining the computational complexity of the nonredundant contourlet transform, this thesis applies the NRCT to fingerprint image compression, since fingerprint images are examples of images with oriented structures. Based on an appropriately designed filter bank structure, the NRCT is easily compatible with the wavelet transform. Hence a new transform is created called the semi-NRCT, which takes the advantages of the directional selectivity of the NRCT and the lower complexity of the wavelet transform. Finally, this thesis proposes a new fingerprint image compression scheme based on the semi-NRCT. The semi-NRCT-based fingerprint image compression is compared with other transform-based compressions, for example the wavelet-based and the contourlet-based algorithms, and is shown to perform favorably.

Actions (login required)

View Item