Image Feature Detection Based on Phase Congruency by Monogenic Filters with New Noise Estimation

摘要

Phase congruency is a relative unknown and powerful image processing technique for segmentation, having been used in diatom image processing, microscopic algae found in water and used to evaluate its quality. However, an important limitation of phase congruence is its sensitivity to noise. To prevent noise from affecting segmentation results, a good noise level estimation is necessary. It can be done with the analysis of the image of local energy. In this paper, we propose the use of the Weibull distribution to estimate the noise profile of the local energy image. The results are compared, in diatom images, with those obtained with the commonly employed Rayleigh distribution and the exponential. The results showed that the Weibull distribution allows a better estimation of the noise level.