【转载】C# OpenCvSharp实现图像校正：自动寻找角点，透视变换轻松搞定！

效果

项目源代码

using OpenCvSharp;
using OpenCvSharp.Extensions;
using System;
using System.Drawing;
using System.Windows.Forms;

namespace OpenCvSharp_图像校正
{
    public partial class Form1 : Form
    {
        public Form1()
        {
            InitializeComponent();
        }

        string img = "test.png";

        private void Form1_Load(object sender, EventArgs e)
        {
            pictureBox1.Image = new Bitmap(img);
        }

        private void button1_Click(object sender, EventArgs e)
        {
            Mat src = new Mat(img);

            //转化为灰度图
            //Cv2.CvtColor(src, src, ColorConversionCodes.RGB2GRAY);

            InputArray kernel = Cv2.GetStructuringElement(MorphShapes.Rect, new OpenCvSharp.Size(3, 3));
            Cv2.MorphologyEx(src, src, MorphTypes.Close, kernel, new OpenCvSharp.Point(-1, -1), 3);
            //Cv2.ImShow("MorphologyEx", src);

            /*
                ksize，高斯内核大小，ksize.width和ksize.height必须是正奇数，两者可以不相同，值越大越模糊
                sigmaX，Y轴方向的标准差，值越大越模糊
                sigmaY，X轴方向的标准差，值越大越模糊
             */
            Cv2.GaussianBlur(src, src, new OpenCvSharp.Size(11, 11), 2, 2);
            //Cv2.ImShow("GaussianBlur", src);

            //Canny边缘检测
            Mat canny_Image = new Mat();
            Cv2.Canny(src, canny_Image, 10, 30, 3, false);

            OpenCvSharp.Point[][] contours;
            HierarchyIndex[] hierarchly;
            /*
           findContours找到轮廓
           第一个参数：单通道图像矩阵，可以是灰度图，但更常用的是二值图像，一般是经过Canny、拉普拉斯等边缘检测算子处理过的二值图像；
           第二个参数：contours 
           第三个参数：hierarchy
           第四个参数：轮廓的检索模式
             取值一：CV_RETR_EXTERNAL 只检测最外围轮廓，包含在外围轮廓内的内围轮廓被忽略
             取值二：CV_RETR_LIST     检测所有的轮廓，包括内围、外围轮廓，但是检测到的轮廓不建立等级关系，彼此之间独立，没有等级关系，这就意味着这个检索模式下不存在父轮廓或内嵌轮廓，所以hierarchy向量内所有元素的第3、第4个分量都会被置为-1，具体下文会讲到
             取值三：CV_RETR_CCOMP    检测所有的轮廓，但所有轮廓只建立两个等级关系，外围为顶层，若外围内的内围轮廓还包含了其他的轮廓信息，则内围内的所有轮廓均归属于顶层
             取值四：CV_RETR_TREE     检测所有轮廓，所有轮廓建立一个等级树结构。外层轮廓包含内层轮廓，内层轮廓还可以继续包含内嵌轮廓。
           第五个参数：轮廓的近似方法
             取值一：CV_CHAIN_APPROX_NONE   保存物体边界上所有连续的轮廓点到contours向量内
             取值二：CV_CHAIN_APPROX_SIMPLE 仅保存轮廓的拐点信息，把所有轮廓拐点处的点保存入contours向量内，拐点与拐点之间直线段上的信息点不予保留
             取值三和四：CV_CHAIN_APPROX_TC89_L1，CV_CHAIN_APPROX_TC89_KCOS使用teh-Chinl chain 近似算法
           第六个参数：Point偏移量，所有的轮廓信息相对于原始图像对应点的偏移量，相当于在每一个检测出的轮廓点上加上该偏移量，且Point可以是负值。不填为默认不偏移Point()
          */
            Cv2.FindContours(canny_Image, out contours, out hierarchly,
                RetrievalModes.External,
                ContourApproximationModes.ApproxSimple,
                new OpenCvSharp.Point(0, 0));

            if (contours.Length == 0)
            {
                MessageBox.Show("边缘检测失败");
                return;
            }

            Random rnd = new Random();
            Scalar color;
            color = new Scalar(0, 255, 0);
            for (int i = 0; i < contours.Length; i++)
            {
                color = new Scalar(rnd.Next(0, 255), rnd.Next(0, 255), rnd.Next(0, 255));
                Cv2.DrawContours(src, contours, i, color, 2, LineTypes.Link4);
            }
            //Cv2.ImShow("contours", src);

            //求出面积最大的轮廓
            double max_area = 0.0;
            double currentArea = 0.0;
            OpenCvSharp.Point[] max_contour = null;
            for (int i = 0; i < contours.Length; i++)
            {
                currentArea = Cv2.ContourArea(contours[i]);
                if (currentArea > max_area)
                {
                    max_area = currentArea;
                    max_contour = contours[i];
                }
            }

            //多边形拟合凸包的四个顶点
            OpenCvSharp.Point[] hull = Cv2.ConvexHull(max_contour);
            double epsilon = 0.02 * Cv2.ArcLength(max_contour, true);
            OpenCvSharp.Point[] approx = Cv2.ApproxPolyDP(hull, epsilon, true);

            if (approx.Length != 4)
            {
                MessageBox.Show("拟合凸包的四个顶点失败");
                return;
            }

            Scalar scalar2 = new Scalar(0, 255, 255);

            Cv2.Line(src, approx[0], approx[1], scalar2, 1, LineTypes.Link4);
            Cv2.Line(src, approx[1], approx[2], scalar2, 1, LineTypes.Link4);
            Cv2.Line(src, approx[2], approx[3], scalar2, 1, LineTypes.Link4);
            Cv2.Line(src, approx[3], approx[0], scalar2, 1, LineTypes.Link4);

            //排序
            Array.Sort(approx, (cs1, cs2) =>
            {
                if (cs1 != null && cs1 != null)
                {
                    if (cs1.Y > cs2.Y)
                        return 1;
                    else if (cs1.Y == cs2.Y)
                    {
                        if (cs1.X < cs2.X)
                            return 1;
                        else return -1;
                    }
                    else
                        return -1;
                }
                return 0;

            });

            //算法找出的角点
            OpenCvSharp.Point2f[] srcPt = new OpenCvSharp.Point2f[4];
            srcPt[0] = approx[0];
            srcPt[1] = approx[1];
            srcPt[2] = approx[3];
            srcPt[3] = approx[2];

            //最小外接矩形
            RotatedRect rect = Cv2.MinAreaRect(srcPt);
            Rect box = rect.BoundingRect();
            OpenCvSharp.Point2f[] dstPt = new OpenCvSharp.Point2f[4];

            dstPt[0].X = box.X;
            dstPt[0].Y = box.Y;

            dstPt[1].X = box.X + box.Width;
            dstPt[1].Y = box.Y;

            dstPt[2].X = box.X + box.Width;
            dstPt[2].Y = box.Y + box.Height;

            dstPt[3].X = box.X;
            dstPt[3].Y = box.Y + box.Height;

            Mat src2 = new Mat(img);
            Mat final = new Mat();
            Mat warpmatrix = Cv2.GetPerspectiveTransform(srcPt, dstPt);//获得变换矩阵
            Cv2.WarpPerspective(src2, final, warpmatrix, src.Size());//投射变换，将结果赋给final

            Bitmap temp = BitmapConverter.ToBitmap(final);

            pictureBox2.Image = temp;

            DrawLine(srcPt, dstPt);

            //Application.DoEvents();
            //System.Threading.Thread.Sleep(1000);
            //pictureBox2.Image = CutImage(temp, (int)p2f[0].X, (int)p2f[0].Y, (int)p2f[2].X, (int)p2f[2].Y);

        }

        void DrawLine(OpenCvSharp.Point2f[] srcPt, OpenCvSharp.Point2f[] dstPt)
        {
            Bitmap bmp = new Bitmap(img);
            Graphics g = Graphics.FromImage(bmp);

            Pen pen = new Pen(Color.Red, 3);
            Pen pen2 = new Pen(Color.Blue, 3);

            g.DrawLine(pen, srcPt[0].X, srcPt[0].Y, srcPt[1].X, srcPt[1].Y);
            g.DrawLine(pen, srcPt[1].X, srcPt[1].Y, srcPt[2].X, srcPt[2].Y);
            g.DrawLine(pen, srcPt[2].X, srcPt[2].Y, srcPt[3].X, srcPt[3].Y);
            g.DrawLine(pen, srcPt[3].X, srcPt[3].Y, srcPt[0].X, srcPt[0].Y);

            g.DrawLine(pen2, dstPt[0].X, dstPt[0].Y, dstPt[1].X, dstPt[1].Y);
            g.DrawLine(pen2, dstPt[1].X, dstPt[1].Y, dstPt[2].X, dstPt[2].Y);
            g.DrawLine(pen2, dstPt[2].X, dstPt[2].Y, dstPt[3].X, dstPt[3].Y);
            g.DrawLine(pen2, dstPt[3].X, dstPt[3].Y, dstPt[0].X, dstPt[0].Y);

            pictureBox1.Image = bmp;

        }

        /// <summary>
        /// 剪裁图片
        /// </summary>
        /// <param name="src">原图片</param>
        /// <param name="left">左坐标</param>
        /// <param name="top">顶部坐标</param>
        /// <param name="right">右坐标</param>
        /// <param name="bottom">底部坐标</param>
        /// <returns>剪裁后的图片</returns>
        public Image CutImage(Image src, int left, int top, int right, int bottom)
        {
            Bitmap srcBitmap = new Bitmap(src);
            int width = right - left;
            int height = bottom - top;
            Bitmap destBitmap = new Bitmap(width, height);
            using (Graphics g = Graphics.FromImage(destBitmap))
            {
                g.Clear(Color.Transparent);
                //设置画布的描绘质量         
                g.CompositingQuality = System.Drawing.Drawing2D.CompositingQuality.HighQuality;
                g.SmoothingMode = System.Drawing.Drawing2D.SmoothingMode.HighQuality;
                g.InterpolationMode = System.Drawing.Drawing2D.InterpolationMode.HighQualityBicubic;
                g.DrawImage(srcBitmap, new Rectangle(0, 0, width, height), left, top, width, height, GraphicsUnit.Pixel);
            }
            return destBitmap;
        }
    }
}

From 公众号:天天代码码天天