Simultaneous Multi-type Feature Separation for Complex Structured Surface Analysis

Manufactured parts with complex structured surfaces have been widely used in automobile, bio-engineering, medical and consumer electronics etc. The content of a complex structured surface is often complex. The organization of surface features, the feature types and sizes all have fundamental effects on the resulting function of the surface. There is not a single transform which is optimal to represent all the contained features. In this research, a novel method called Morphological Component Analysis (MCA) is investigated to characterise different types of features on complex structured surfaces. Due to the fact that different types of features on the structured surface present different morphological aspects, the MCA method employs a set of dictionaries/basis functions with various morphological characteristics to represent various types of features, such as random noise, directional texture, steps, edges and others. By using MCA one can separate different types of features simultaneously. It is recommended that the wavelet type functions, especially DT-CWT, can be selected for modelling the geometrical features (such as peaks, valley, holes, steps and edges); the Fourier type basis functions such as basis function of DCT can be used for mapping of the surface texture. Other advantages of the MCA include: it can separate multiple types of features simultaneously and have better feature preservation properties compared with the convolution based filters.