MATLAB 亚像素点检测代码；

最新推荐文章于 2026-05-01 06:46:28 发布

原创

最新推荐文章于 2026-05-01 06:46:28 发布 · 3.8k 阅读

标签

#matlab #buffer #parameters #function #c

收录于

该函数用于在MATLAB中精炼图像角点的位置，采用最小二乘迭代方法提高角点检测的亚像素精度。代码首先定义了窗口大小、误差阈值和最大迭代次数，然后通过创建掩模矩阵并进行子像素数据采样、梯度计算及位置更新，不断迭代优化角点坐标，直至达到预设的精度标准。

用法同 openCV, 使用最小二乘迭代

function corners_tuned = refine_pos( src, corners) % corners: N*2
%UNTITLED Summary of this function goes here
% Detailed explanation goes here
corners_tuned = corners;

% some parameters needed to set;
win.width = 3;
win.height = 3;
eps = 0.00001;
max_iters = 20;
count = size(corners, 1);

% parameters derived;
win_w = win.width*2 + 1;
win_h = win.width*2 + 1;
coeff = 1.0/(win.width*win.width);
maskX = zeros(win_w, 1, 'double');
maskY = zeros(win_h, 1, 'double');
mask = zeros(win_h, win_w, 'double');

for i = -win.width:1:win.width
maskX(i + win.width + 1) = exp(-i*i*coeff);
end
maskY = maskX;

for i = 1:1:win_w
    for j = 1:1:win_h
        mask(i, j) = maskX(j)*maskY(i);
    end
end

for pt_i = 1:1:count
    ct = corners(pt_i, :);
    iter = 0;
    err = 0.0;

    nci.x = ct(1, 2);
    nci.y = ct(1, 1);

    str = sprintf('============== p%d iteration =============', pt_i);
    disp(str);

    while (1)
        % s1: get the sub-pixsel data;
  &