Stitching Illumination-Matched Maps: A LIMA-Based Pipeline for Robust Reference Map Generation

Mapping, navigation, and landing in high-latitude regions, such as the lunar south pole, are challenged by extreme lighting variability. Conventional reference maps often fail to match the sun elevation and azimuth conditions encountered during operations, deteriorating feature matching and terrain-relative navigation. We propose a novel approach to synthesize illumination-consistent reference maps from a set of images captured under diverse lighting conditions. Our method extends the Lighting-Invariant Matching Algorithm (LIMA) to full-area mosaic stitching by selecting locally optimal image pairs for each tile, creating inverse composite image patches under a user-defined target sun vector, and blending them using a priority-based stitching strategy. We evaluate the approach on a 29-image Lunar Reconnaissance Orbiter (LRO) Narrow Angle Camera (NAC) image dataset and demonstrate that the constructed mosaics better match a held-out ground truth image than any individual input frame. Fast Fourier Transform (FFT)-based phase correlation and Scale Invariant Feature Transform (SIFT) inlier ratios confirm improved alignment under both normal and azimuth-restricted input conditions. The result is a spatially coherent, illumination-matched reference map applicable to planetary missions where operational lighting differs from archival imagery.