Dynamic Incorporation ofWavelet Filter in Fuzzy C-Means for Efficient and Noise-Insensitive MR Image Segmentation

AbstractImage intensity in magnetic resonance (MR) images in the presence of noise obeys Rician distribution. The signal-dependent Rician noise makes accurate image segmentation a challenging task. Although existing fuzzy c-means (FCM) variants with local filters improve the segmentation performance, they are less effective for reducing the negative effect from Rician noise, and the repeatedly applied filter increases their computational intensiveness. To address this issue, we propose a novel image segmentation method which dynamically incorporates wavelet-based noise detector and filter in the FCM membership function. The modified algorithm is designed to exploit both frequency and spatial information in the images and minimizes clustering errors caused by Rician noise. Furthermore, efficiency of the proposed method can be enhanced by the strategy of applying filter only when noise is detected. The experimental results of segmentation on synthetic and brain MR images, demonstrate the computational effic...