手写Findhomography，能否考虑引入ocvm中

[futz12]

手写过Findhomography（在不显著增加体积情况下，增加一些功能），个人感觉比较有用。能否考虑引入ocvm中

// 对OpenCV的findHomography的重写，利用SVD求解
    cv::Mat Findhomography(std::vector<cv::Point2f> src, std::vector<cv::Point2f> target) {
        // 求解系统为 A*X - B = 0

        const int n = src.size();
        float x[9] = {0};

        cv::Mat A = cv::Mat::zeros(2 * n, 8, CV_32F);
        cv::Mat B = cv::Mat::zeros(2 * n, 1, CV_32F);
        cv::Mat X(8, 1, CV_32F, x);

        for (int i = 0; i < n; i++) {
            A.at<float>(i, 0) = src[i].x;
            A.at<float>(i + n, 3) = src[i].x;
            A.at<float>(i, 1) = src[i].y;
            A.at<float>(i + n, 4) = src[i].y;

            A.at<float>(i, 2) = 1;
            A.at<float>(i + n, 5) = 1;

            A.at<float>(i, 6) = -src[i].x * target[i].x;
            A.at<float>(i, 7) = -src[i].y * target[i].x;

            A.at<float>(i + n, 6) = -src[i].x * target[i].y;
            A.at<float>(i + n, 7) = -src[i].y * target[i].y;

            B.at<float>(i, 0) = target[i].x;
            B.at<float>(i + n, 0) = target[i].y;
        }

        cv::solve(A, B, X, cv::DECOMP_SVD);


        x[8] = 1;
        cv::Mat H = cv::Mat(3, 3, CV_32F, x).clone();
        return H;
    }
···

[nihui]

如果你的实现和标准 opencv findHomography 行为保持一致，欢迎 pull request