Multi-Contrast MRI Denoising via Tissue Segmentation

The patent applies different filtering strategies to different regions of the image based on tissue type identification. By segmenting the image into distinct tissue regions (e.g., gray matter, white matter, CSF) and applying appropriate filters to each region, the method achieves effective noise reduction while preserving local image clarity and boundary definitions. This localized approach prevents the excessive blurring that occurs with global low-pass filtering.

The patent dynamically adjusts filtering parameters based on the identified tissue type and local image characteristics. By changing filter parameters (such as kernel size, threshold values, and smoothing strength) according to the specific tissue region being processed, the method achieves effective noise reduction while maintaining the natural appearance of the image and preserving detailed features. This adaptive parameter adjustment prevents the unnatural look and feature erosion associated with fixed non-linear filters.

The patent segments the MRI image into distinct tissue regions based on contrast characteristics and luminance value patterns. By identifying and separating different tissue types (gray matter, white matter, cerebrospinal fluid, and other structures), the method can apply optimized denoising parameters to each segment, thereby reducing noise while preserving the detailed features and natural appearance critical for accurate diagnostic analysis.