如何在emgu-cv c＃中将遮罩设置为GrabCut方法？

Posted 2023-02-15

我正在尝试在emgu cv中使用GrabCut方法

这是我的代码

 Matrix<double> bg = new Matrix<double>(1, 65);

        bg.SetZero();



        Matrix<double> fg = new Matrix<double>(1, 65);
        fg.SetZero();


        Image<Gray, byte> mask =  new Image<Gray, byte>(img.Size);


        Rectangle rect = new Rectangle(img.Cols / 4, 0, (int)((double)img.Width / (0.75)), img.Height );


        CvInvoke.GrabCut(img, mask, rect,
           bg, fg, 5, Emgu.CV.CvEnum.GrabcutInitType.InitWithRect);


        for (int x = 0; x < mask.Cols; x++)
        
            for (int y = 0; y < mask.Rows; y++)
            
                if (mask[y, x].Intensity == new Gray(1).Intensity || mask[y, x].Intensity == new Gray(3).Intensity)
                
                    mask[y, x] = new Gray(1);
                
                else
                
                    mask[y, x] = new Gray(0);
                
            

        

        img = img.Mul(mask.Convert<Bgr,byte>());




        imageBox3.Image = img;

我的img：

结果：

但是我想要T恤，所以我尝试使用这个面具（我在photoshop中制作了）

mask：

我改变了面具，所以我的代码变成了这样：

 Matrix<double> bg = new Matrix<double>(1, 65);

        bg.SetZero();



        Matrix<double> fg = new Matrix<double>(1, 65);
        fg.SetZero();


        Image<Gray, byte> mask =  new Image<Gray, byte>(@"C:\Users\iP\Desktop\exaples\mas.jpg");

        //here i set the only white pixels (foreground object ) to 1 and 0 for else
        for (int x = 0; x < mask.Cols; x++)
        
            for (int y = 0; y < mask.Rows; y++)
            
                if (mask[y, x].Intensity > new Gray(200).Intensity)
                
                    mask[y, x] = new Gray(1);
                
                else
                
                    mask[y, x] = new Gray(0);
                
            
        


        Rectangle rect = new Rectangle(img.Cols / 4, 0, (int)((double)img.Width / (0.75)), img.Height );


        CvInvoke.GrabCut(img, mask, rect,
           bg, fg, 5, Emgu.CV.CvEnum.GrabcutInitType.InitWithRect);


        for (int x = 0; x < mask.Cols; x++)
        
            for (int y = 0; y < mask.Rows; y++)
            
                if (mask[y, x].Intensity == new Gray(1).Intensity || mask[y, x].Intensity == new Gray(3).Intensity)
                
                    mask[y, x] = new Gray(1);
                
                else
                
                    mask[y, x] = new Gray(0);
                
            

        

        img = img.Mul(mask.Convert<Bgr,byte>());




        CvInvoke.Imshow("result", img);

但是我用第一个口罩得到了相同的结果（没有T恤）

我的代码中的错误在哪里？

并且我尝试将Emgu.CV.CvEnum.GrabcutInitType.InitWithRect更改为Emgu.CV.CvEnum.GrabcutInitType.InitWithMask

我知道了

答案

此代码现在正在运行:)

Matrix<double> bg = new Matrix<double>(1, 65);

        bg.SetZero();



        Matrix<double> fg = new Matrix<double>(1, 65);
        fg.SetZero();


        Image<Gray, byte> mask = new Image<Gray, byte>(img.Size);


        Rectangle rect = new Rectangle(img.Cols / 4, 0, (int)((double)img.Width / (0.75)), img.Height );


        CvInvoke.GrabCut(img, mask, rect,
           bg, fg, 5, Emgu.CV.CvEnum.GrabcutInitType.InitWithRect);


        Image<Gray, byte> mask2 = new Image<Gray, byte>(@"C:\Users\iP\Desktop\exaples\mas.jpg");

        ////here i set the only white pixels (foreground object ) to 1 and 0 for else
        for (int x = 0; x < mask.Cols; x++)
        
            for (int y = 0; y < mask.Rows; y++)
            
                if (mask2[y, x].Intensity > new Gray(200).Intensity)
                
                    mask[y, x] = new Gray(1);
                
                else
                

                
            
        

        CvInvoke.GrabCut(img, mask, rect,
             bg, fg, 5, Emgu.CV.CvEnum.GrabcutInitType.InitWithMask);


        for (int x = 0; x < mask.Cols; x++)
        
            for (int y = 0; y < mask.Rows; y++)
            
                if (mask[y, x].Intensity == new Gray(1).Intensity || mask[y, x].Intensity == new Gray(3).Intensity)
                
                    mask[y, x] = new Gray(1);
                
                else
                
                    mask[y, x] = new Gray(0);
                
            

        

        img = img.Mul(mask.Convert<Bgr,byte>());




        CvInvoke.Imshow("result", img);

结果：

视频：https://www.youtube.com/watch?v=463TMas4vHU