Blending Overlapping Images Using Iterative Gradient Averaging

The patent transforms the blending problem from distance-based scalar values to gradient-based vector fields. By computing gradient vectors at each pixel and iteratively averaging neighboring gradients, the method changes the parameter space from simple distance metrics to continuous gradient fields, eliminating triangular artifacts and first-order discontinuities while maintaining smooth edge transitions.

The patent employs an iterative dynamic process where gradient vectors are repeatedly updated by averaging with neighboring pixels until convergence. This dynamic iteration allows the blend region to naturally adapt to arbitrary projector configurations without fixed geometric assumptions, preventing artifacts while achieving smooth continuous blending across the entire overlap area.

The patent changes the fundamental parameters from distance-based scalar blending to gradient-based vector field blending. This parameter transformation enables the algorithm to handle arbitrary projector configurations smoothly by computing and averaging gradient vectors, eliminating the triangular artifacts and discontinuities that plague traditional distance-based methods while maintaining full adaptability to any projector arrangement.

The patent ensures continuous blending action across the entire overlap region by iteratively propagating gradient information from all boundaries simultaneously. This continuous iterative process eliminates discrete artifacts and maintains smooth transitions throughout the blend region, regardless of projector configuration, achieving both versatility and high visual quality.