A Dynamic Programming Stereo Matching Method Combining Adaptive Weighting

Abstract

Traditional stereo matching algorithms based on dynamic programming only consider the disparity smoothing constraint
of adjacent pixels in the row direction of the image, but ignore the disparity smoothing constraint of neighboring pixels in the column
direction, resulting in a solution also known as the epipolar optimal solution.This matching method ignores the smoothness constraint in
the column direction, resulting in a high mismatch rate in the column direction, resulting in clear disparity stripes on the resulting disparity
map.In response to this issue, this article proposes a stereo matching algorithm based on row row bidirectional dynamic programming,
which improves the matching eff ect to a certain extent.In the process of minimizing the energy function, the dynamic optimization method
is fi rst used in the row direction to provide the energy minimization value of the disparity map. Then, based on the solution result of the row
dynamic programming, the corresponding data items in the energy function are updated. Finally, the dynamic programming optimization is
performed in the column direction to obtain a dense disparity map.