Geometry-Based Field Prediction for MRI Phase Artifact Removal

The image space is divided into multiple subspaces based on dominant field directions. Each subspace is processed independently with a tailored high pass filter that preserves local phase information while removing aliasing artifacts specific to that region's field orientation.

Different high pass filtering strategies are applied to different spatial regions according to their local field characteristics. The filtering strength and direction are adapted locally rather than applying a uniform filter across the entire image, preserving local phase information while removing aliasing.

The processing is segmented into direction-specific subspaces, each handled independently. This breaks down the complex global polynomial fitting problem into simpler local filtering operations that are computationally more efficient and easier to implement.

The mechanical process of phase unwrapping followed by polynomial fitting is replaced with a direct high pass filtering approach in the spatial domain, operating on complex image data without requiring phase unwrapping. This substitution simplifies the processing pipeline while achieving similar background field removal.

A universal high pass filtering approach is applied across all slices and directions using a consistent algorithmic framework. This multi-functional approach handles different field orientations and slice configurations with a single unified method, improving processing efficiency while maintaining effectiveness.

Instead of performing exhaustive polynomial fitting with variable order polynomials across all slices, the method applies a standardized high pass filter that provides sufficient aliasing removal with reduced computational overhead, accepting a slight approximation for significant speed improvement.