Improved bi-dimensional EMD and Hilbert spectrum for the analysis of textures
Pattern Recognition
The complex bidimensional empirical mode decomposition
Signal Processing
On analysis of bi-dimensional component decomposition via BEMD
Pattern Recognition
| Hi-index | 0.00 |
A novel decomposition approach called Neighborhood Limited Empirical Mode Decomposition (NLEMD) is presented. Unlike other EMD approaches, only local neighborhood data are used to estimate the local mean value in NLEMD. It is surprising that NLEMD has inherited all advantages from EMD proposed by Huang. Meanwhile, some inherent bugs in EMD have been avoided. In NLEMD, the signal can be decomposed into several basic components, while the minimum frequency existing in Intrinsic Mode Functions (IMFs) can be controlled in each decomposition step. The experiments show that NLEMD works well for the 1D signal and 2D signal. The decomposition results of images are distinct and precise, and no gray spots appear compared with ones from the existing algorithms. NLEMD is one novel multi-resolution approach, and every IMF has different time and frequency resolution . By extracting useful data from every IMF and fusing them, we can obtain better image visually. In this paper, NLEMD has succeeded in high dynamic range image compression, image fusion etc.