Voxel-wise Weighted MR Image Enhancement using an Extended Neighborhood Filter
This work addresses image enhancement for medical applications like multiple sclerosis and stroke, but it appears incremental as it builds on existing filtering methods.
The paper tackles the problem of enhancing features in medical images with narrow or distributed ROIs, such as in multiple sclerosis or angiograms, by proposing an edge-preserving denoising filter that uses extended neighborhood directions, and it demonstrates advantages over specialized non-linear filters using numerical phantom data and performance evaluation on simulated and clinical images.
We present an edge preserving and denoising filter for enhancing the features in images, which contain an ROI having a narrow spatial extent. Typical examples include angiograms, or ROI spatially distributed in multiple locations and contained within an outlying region, such as in multiple-sclerosis. The filtering involves determination of multiplicative weights in the spatial domain using an extended set of neighborhood directions. Equivalently, the filtering operation may be interpreted as a combination of directional filters in the frequency domain, with selective weighting for spatial frequencies contained within each direction. The advantages of the proposed filter in comparison to specialized non-linear filters, which operate on diffusion principle, are illustrated using numerical phantom data. The performance evaluation is carried out on simulated images from BrainWeb database for multiple-sclerosis, acute ischemic stroke using clinically acquired FLAIR images and MR angiograms.