科研之友
微信
新浪微博
Facebook
分享链接
摘要
The nonlocal maximum-likelihood (NLML) filter is considered to be a denoising method, which preserves detail information effectively in magnetic resonance (MR) images. However, one drawback of the existing nonlocal maximum-likelihood (NLML) denoising method is that the similarity measure is based on the similarity of the intensity of pixels in the neighborhood. The measure of similarity is conducted in the discrete cosine transform (DCT) subspace of pixel neighborhood in this work. Due to low data correlation and high energy compaction of DCT, the accuracy of the selection of similarity samples is improved. The other drawback is that we use a fixed sample size for the ML estimation. In this work, the samples are adaptively selected by using K-means clustering algorithm in the DCT subspace of pixel neighborhood. Qualitative and quantitative comparison experiments among the proposed filter, nonlocal maximum-likelihood filter, nonlocal means filter, and variants of these filters are carried out on MR images. The comparison experiments indicate that the proposed filter achieves better capability of removing noise while preserving image details in MR images.
