A clustering framework based on FCM and texture features for denoising ESPI fringe patterns with variable density

Abstract

Denoising of electronic speckle pattern interferometry (ESPI) fringe patterns with variable density is challenging. In this paper, we present a method to denoise ESPI fringe patterns with variable density by constructing a clustering framework. The framework first clusters the variable density fringes into low density fringes and high density fringes according to the texture features of fringes by using fuzzy c-means (FCM) clustering algorithm. The texture features include energy, contrast, homogeneity and correlation, which can characterize the densities of fringes. Then, the effective existing methods are adopted to denoise the low and high density fringes individually. Finally, the filtered low and high density fringes are combined to obtain the final denoising results of the fringe patterns with variable density. With the help of proposed framework, variable density fringes can be clustered into low and high density fringes without calculating the density map of fringe pattern. Furthermore, the filtering strength applied in high density fringes can be different from that in low density fringes in a fringe pattern, which can improve the performance of any filtering method in the denosing of fringe patterns with variable density. Experiments on computer-simulated and experimentally obtained ESPI fringe patterns tested with several filtering methods show that the proposed framework provides consistently better results than denoising the image directly, in terms of signal-to-noise ratio, equivalent number of looks and edge preservation metrics.