A novel multi-scale edge detection technique based on wavelet analysis with application in multiphase flows

A novel non-linear weighted statistical algorithm for multi-scale edge detection of gray image based on wavelet analysis is proposed to extract interface in multiphase flows. In this method, local modulus maximum of gray gradient along the phase angle direction is regarded as the possible edge (interface) of an image. A two-dimensional discrete Gaussian function with zero average value is adopted as smoothening filter. The influences of filter length, scale and threshold on the extracted edges are discussed in detail. In order to realize multi-scale detection, the local maximums of the wavelet transform modulus calculated under a series of the wavelet transform scales are used to detect the possible edge (interface) of an image. A modified Gaussian function is suggested to weight non-linearly those possible edges (interfaces) detected on different scales and a statistical function is proposed to synthesize the possibility of each pixel being edge over all scales. Compared with Canny operator, this new algorithm can extract the edge perfectly even with strong noise and light reflection, which shows the potential to extract the interface in multiphase flows.

A novel non-linear weighted statistical algorithm based on multi-scale wavelet transform with modified Gaussian filter is proposed to extract interface in multiphase flow. In this algorithm, local maximum of the gray gradient module of each pixel in the image is used to detect the interface. When applied to multiphase flow, this novel technique extracted the shape of different particles accurately.Figure optionsDownload as PowerPoint slide