Multifractal Image Texture Analysis Combined with 2D Empirical Mode Decomposition

The surface texture and microstructure of digital images have an important influence on the construction of features such as image analysis, transformation, and compression. Studies have shown that the fractal spectrum parameters of different types of subject matter will be significantly different. Multifractal spectra and scaling indices quantify the heterogeneity of structural features, demonstrating multiscaling properties. This paper proposes a multifractal spectrum algorithm combining empirical mode decomposition (EMD) and wavelet leaders, starting from the image texture classification task. This method describes the surface shape and microstructure of the image, extends the mode decomposition of the one-dimensional signal in the Hilbert-Huang transform to the two-dimensional image, and gives an image descriptor based on the fractal spectrum. Simulation results demonstrate the accuracy of the proposed method.