Capture desktop , Make it 256 color and send it over internet

Question

I am working on a project and in this project I need to capture desktop screen and send it over internet to other client. To compress the image I want to convert it to a 256 color image. I have a general 256 color palette and I am using Euclidean distance to find the nearest color. The problem is that I need to send these images in 10-15 frames per second and making a 256 color image takes 7 seconds. I am wondering how other programs (like teamviewer or real VNC ) are doing this.

for (int y=0;y<900;y++) //loop through the height
     for (int x=0;x<1600;x++) //loop through the width
         for (int p=0;p<256;p++) //loop through palette colors
             {
                calculate Euclidean distance of each pixel for each color in pallette and 
                 find the nearest color
                 ** these nested loops take 7 seconds to complete
             }

Thanks

Answer 1

OK. After a few days of struggling with many capturing methods and color quantiziers I finally found a solution. Now I am able to send whole desktop image at 10~14 FPS and changed regions of desktop at 20~30 FPS.

In the code I used rcravens's classes to capture screen and changes of screen. Then I cropped the image to 10 samll pieces . After that I took small pieces and made them 256-color using an Octree Color Quantizier explained here thanks to @Dai who pointed me to that direction. After color reduction I converted each piece to a byte array and compressed them using LZ4.Net library.

Here is the code:

    int all_count = 0;
    Bitmap _desktop = null;

    Bitmap _merged_bitmap = new Bitmap(1600, 900);

    int _height_part_ = 0;
    int _total_rows = 10;
    Bitmap[] crops = null;
    Bitmap[] _new_crops = null;
    Stopwatch sw = new Stopwatch();

    int _desktop_height = 0;
    int _desktop_width = 0;

    ImageManipulation.OctreeQuantizer _q ;
    RLC.RemoteDesktop.ScreenCapture cap = new RLC.RemoteDesktop.ScreenCapture();



   private void CaptureAndSend()
    {
        sw.Restart();

        //cap = new RLC.RemoteDesktop.ScreenCapture();

        int _left = -1, _top = -1; //Changed regions
        _desktop = cap.Screen(out _left, out _top); //Capture desktop or changed region of it

        if (_desktop == null) return; //if nothing has changed since last capture skip everything

        _desktop_height = _desktop.Height;
        _desktop_width = _desktop.Width;

        // If very small part has changed since last capture skip everything
        if (_desktop_height < 10 || _desktop_width < 10) return; 

        TotalRows(_total_rows); // Calculate the total number of rows 

        crops = new Bitmap[_total_rows]; // Cropped pieces of image
        _new_crops = new Bitmap[_total_rows];

        for (int i = 0; i < _total_rows - 1; i++) //Take whole image and split it into smaller images
            crops[i] = CropRow(i);
        crops[_total_rows - 1] = CropLastRow(_total_rows - 1);


        Parallel.For(0, _total_rows, i =>
        {
            ImageManipulation.OctreeQuantizer _q = new ImageManipulation.OctreeQuantizer(255, 4); // Initialize Octree
            _new_crops[i] = _q.Quantize(crops[i]);

            using (MemoryStream ms=new MemoryStream())
            { 
                _new_crops[i].Save(ms, ImageFormat.Png);
                //Install-Package LZ4.net
                //Compress each part and send them over network
                byte[] data = Lz4Net.Lz4.CompressBytes(ms.ToArray(), Lz4Net.Lz4Mode.HighCompression);

                all_count += data.Length; //Just to check the final size of image
            }                  
        });



        Console.WriteLine(String.Format("{0:0.0} FPS , {1} seconds , size {2} kb", 1.0 / sw.Elapsed.TotalSeconds, sw.Elapsed.TotalSeconds.ToString(), all_count / 1024));
        all_count = 0;

    }
    private void TotalRows(int parts)
    {
        _height_part_ = _desktop_height / parts;
    }
    private Bitmap CropRow(int row)
    {
        return Crop(_desktop, new Rectangle(0, row * _height_part_, _desktop_width, _height_part_));
    }
    private Bitmap CropLastRow(int row)
    {
        return Crop(_desktop, new Rectangle(0, row * _height_part_, _desktop_width, _desktop_height - (row * _height_part_)));
    }
 [DllImport("msvcrt.dll", CallingConvention = CallingConvention.Cdecl)]
    private unsafe static extern int memcpy(byte* dest, byte* src, long count);

    private unsafe Bitmap Crop(Bitmap srcImg, Rectangle rectangle)
    {
        if ((srcImg.Width == rectangle.Width) && (srcImg.Height == rectangle.Height))
            return srcImg;

        var srcImgBitmapData = srcImg.LockBits(new Rectangle(0, 0, srcImg.Width, srcImg.Height), ImageLockMode.ReadOnly, srcImg.PixelFormat);
        var bpp = srcImgBitmapData.Stride / srcImgBitmapData.Width; // 3 or 4
        var srcPtr = (byte*)srcImgBitmapData.Scan0.ToPointer() + rectangle.Y * srcImgBitmapData.Stride + rectangle.X * bpp;
        var srcStride = srcImgBitmapData.Stride;

        var dstImg = new Bitmap(rectangle.Width, rectangle.Height, srcImg.PixelFormat);
        var dstImgBitmapData = dstImg.LockBits(new Rectangle(0, 0, dstImg.Width, dstImg.Height), ImageLockMode.WriteOnly, dstImg.PixelFormat);
        var dstPtr = (byte*)dstImgBitmapData.Scan0.ToPointer();
        var dstStride = dstImgBitmapData.Stride;

        for (int y = 0; y < rectangle.Height; y++)
        {
            memcpy(dstPtr, srcPtr, dstStride);
            srcPtr += srcStride;
            dstPtr += dstStride;
        }

        srcImg.UnlockBits(srcImgBitmapData);
        dstImg.UnlockBits(dstImgBitmapData);
        return dstImg;
    }

I know that my code isn't memory efficient . I'd appreciate it if anyone can help me optimize this code. Thanks again to my friends A. Abramov , Dai , HansPassant , TaW and others.