Nonsubsampled Directional Filter Banks Research Articles

A Novel Intelligent Technique for Recognition of Coke Optical Texture

The classification and recognition of the coke optical texture is one of the key elements to determine the quality and guide the production of cokes. Since the traditional methods are not so ideal in the spatial and frequency domains, a novel algorithm is proposed in this paper to bridge the gap. Firstly, the coke micrograph is decomposed by a new contourlet packet (CP) for multi-scale and multi-direction, which introduces a nonsubsampled wavelet transform and a nonsubsampled directional filter banks (NSDFB). Furthermore, an adaptively weighted 2-directonal 2-dimension PCA method is put forward to extract the feature, which not only reduce the data dimensions, but also help to obtain a set of optimal basis in decomposed sub-bands. Finally, the classes of optical texture in coke micrograph are identified according to the knowledge-based similarity measure criteria on eigenvectors of selected basis. The experimental results indicate that the proposed scheme has a higher and more stable recognition rate than the conventional methods.

复轮廓波包的构造及其图像去噪应用

Considering the normal contourlet transform only decomposed the low frequency coefficients of the signals, and ignored the high frequency coefficients, a novel complex contourlet packet transform is constructed by combining the analytic dual-tree complex wavelet packet transform and nonsubsampled directional filter banks. Then a complex contourlet packet image denoising algorithm based on neighbouring thresholding classification is proposed. The new transform has good characteristics of multiresolution, localization, directionality and anisotropy, as well as translation invariance. Furthermore, it has more abundant direction components. The experimental result shows that the complex contourlet packet transform can restrain Gibbs-like artificial around edges in the course of denoising, and preserve more details and textures of the images efficiently. The PSNR and the visual quality of this algorithm are also superior to the traditional methods.

Image fusion based on a new contourlet packet

Contourlet is a “true” two-dimensional transform that can capture the intrinsic geometrical structure and has been applied to many tasks in image processing. In this paper, a new contourlet packet (CP) is constructed based on a complete wavelet quadtree followed by a nonsubsampled directional filter bank (NSDFB). By combing the finer approximation characteristic of wavelet packet (WP) with the invertible characteristic of NSDFB, the proposed CP has more accurate reconstruction of images than WP. Moreover, the wavelet quadtree decomposition is implemented by the stationary wavelet transform (SWT), so the CP proves to be characteristic of shift-invariant and linear phase by choosing appropriate filters. After the proposed CP transform on the fusing images, a pulse coupled neural network (PCNN) is used to make a fusion decision, which can obtain better visual result for the global features of the original images being extracted by the output pulses of the PCNN neurons. We compare the performance of our proposed method in image fusion with that of wavelet, contourlet, wavelet packet and other contourlet packet based approaches. The experiment results show the superiorities of the method to its counterparts in image clarity and some numerical guidelines.

The Nonsubsampled Contourlet Transform: Theory, Design, and Applications

In this paper, we develop the nonsubsampled contourlet transform (NSCT) and study its applications. The construction proposed in this paper is based on a nonsubsampled pyramid structure and nonsubsampled directional filter banks. The result is a flexible multiscale, multidirection, and shift-invariant image decomposition that can be efficiently implemented via the à trous algorithm. At the core of the proposed scheme is the nonseparable two-channel nonsubsampled filter bank (NSFB). We exploit the less stringent design condition of the NSFB to design filters that lead to a NSCT with better frequency selectivity and regularity when compared to the contourlet transform. We propose a design framework based on the mapping approach, that allows for a fast implementation based on a lifting or ladder structure, and only uses one-dimensional filtering in some cases. In addition, our design ensures that the corresponding frame elements are regular, symmetric, and the frame is close to a tight one. We assess the performance of the NSCT in image denoising and enhancement applications. In both applications the NSCT compares favorably to other existing methods in the literature.