Multi-hypothesis spatially-variant autofocus system and method for focusing SAR imagery

A multi-hypothesis spatially-variant autofocus system for processing radar images, the system comprising:

• processing circuitry;
• and memory, the processing circuitry configured to perform an iterative process for a plurality of tile size iterations, wherein to perform the iterative process, the processing circuitry is configured to:
• extract overlapping tiles of an initial tile size from an initial radar image,
• compute an initial figure of merit for each tile;
• apply one or more window size weighting functions to each tile;
• apply one or more phase-gradient autofocus (PGA) algorithms to each window-size weighted version of each tile;
• determine a new figure of merit for each window size weighted version of the tiles for each of the one or more applied PGA algorithms;
• select a correction vector for each tile based on a comparison of the figures of merit;
• apply the selected correction vector to each tile to generate focused tiles;
• and
• update the image by inserting non-overlapping portions of the focused tiles into the image,
• wherein the memory is configured to store the selected correction vectors,
• wherein for each subsequent tile-size iteration, the processing circuitry is configured to generate an updated image by progressively reducing a tile size of the overlapping tiles, and
• wherein for each subsequent tile-size iteration, overlapping tiles with a progressively reduced tile size are extracted from the updated image that is generated from a prior tile-size iteration and an initial figure of merit for each tile is generated for the updated image,
• wherein the system further comprise a multi-hypothesis PGA engine configured for performing a plurality of different PGA algorithms, and
• wherein for each iteration the processing circuitry is configured to apply at least two of the different PGA algorithms to each window-size weighted version of each tile by injecting each window size weighted version of the tiles into the multi-hypothesis PGA engine,
• wherein the plurality of different PGA algorithms includes at least some of:
• a standard PGA weighting algorithm, a coherence-weighted PGA algorithm, a Super-PGA algorithm, and custom PGA weighting algorithm,
• wherein the initial figure of merit is an initial tile contrast,
• wherein the new figure of merit is a new tile contrast, and
• wherein for each window size weighted version of a tile and for each of the applied PGA algorithms, the correction vector that is selected for a tile is selected to provide a highest contrast.

A non-transitory computer-readable storage medium that stores instructions for execution by processing circuitry of a multi-hypothesis spatially-variant autofocus system for processing radar images, the processing circuitry configured to perform an iterative process for a plurality of tile size iterations, wherein to perform the iterative process, the processing circuitry is configured to:

• spatially-segment an initial radar image by extracting a plurality of overlapping image tiles of an initial tile size from the initial radar image, each overlapping image tile being a spatially segmented sub-image of the initial radar image;
• compute an initial figure of merit for each overlapping image tile;
• apply one or more window size weighting functions to each overlapping image tile;
• apply more than one phase-gradient autofocus (PGA) algorithms to each window-size weighted version of each overlapping image tile;
• determine a new figure of merit for each window size weighted version of the tiles for each of the applied PGA algorithms;
• select, on a tile-by-tile basis, a correction vector for each overlapping image tile based on a comparison of the figures of merit, the figures of merit representing contrast metrics for each tile;
• apply, on the tile-by-tile basis, the selected correction vector to each overlapping image tile to generate focused tiles;
• and
• update the image by inserting non-overlapping portions of the focused tiles into the image, wherein the correction vector is selectable based on the figures of merit of different of the applied PGA algorithms.

A method for focusing synthetic aperture radar (SAR) imagery, the method comprising:

• computing phase corrections and applying multiple autofocus strategies to overlapping image tiles for progressively smaller image tile sizes;
• and
• selecting correction factors for the image tiles on a per-tile basis based on various metrics,
• wherein each overlapping image tile is a spatially segmented sub-image of an initial radar image.