Noisy image segmentation model based on the fitting distance energy

Noise often causes considerable difficulties in both gray and color image segmentation. To tackle this problem, we propose a novel region-based level set method, which consists of a curve and two fitting squared distances between the local and global average intensities of the regions. In addition, the model in variational level set formulation consists of a length term that smoothes the zero level set, and a penalty term for regularization. By taking the local intensity means rather than each pixel into consideration, our model possesses the capability of segmenting various severe noisy images. The proposed model has been applied to both synthetic and real images with promising results. © 2014 Binary Information Press.