An improved FCM method for image segmentation based on wavelet transform and particle swarm

Fuzzy C-Means (FCM) algorithm is one of the most commonly used image segmentation algorithms. It has the advantages of unsupervised, easy calculation, soft segmentation and so on. However, for the image containing noise, it will be more obviously disturbed. At the same time, it is sensitive to the initial value and easy to fall into the local minimum. Aiming at solving above problems, a new FCM algorithm is proposed, which combines wavelet transform and improved FCM algorithm. Firstly, the high frequency and low frequency coefficients of different scales are obtained by using the wavelet transform to decompose the image. The Anisotropic Diffusion is used to denoise the decomposed high frequency coefficients. Then, the processed coefficients are reconstructed by wavelet to get the processed images. Finally, the particle swarm optimization algorithm is used to update the FCM cluster centers to get the global optimal value. The experimental results show that the proposed algorithm can better suppress the influence of noise and has better robustness.

I. Introduction

Image segmentation refers to dividing an image into geometric features and non-overlapping regions according to certain criteria. Currently, image segmentation methods mainly include: threshold-based segmentation method [1], particle swarm optimization-based multi-level threshold segmentation [2], edge detection-based segmentation methods, such as Laplacian operator [3], structured forests. [4]; Segmentation based on regional growth and splitting [5], a combination of level set and region growth [6]; a segmentation method combined with specific theory, such as watershed-based labeling [7], deep convolution Neural network method[8].