几种图像处理类库的比较

作者：王先荣

原文；http://www.cnblogs.com/xrwang/archive/2010/01/26/TheComparisonOfImageProcessingLibraries.html

前言

近期需要做一些图像处理方面的学习和研究，首要任务就是选择一套合适的图像处理类库。目前较知名且功能完善的图像处理类库有OpenCv、EmguCv、AForge.net等等。本文将从许可协议、下载、安装、文档资料、易用性、性能等方面对这些类库进行比较，然后给出选择建议，当然也包括我自己的选择。

 

许可协议

类库	许可协议	许可协议网址	大致介绍
OpenCv	BSD	www.opensource.org/licenses/bsd-license.html	在保留原来BSD协议声明的前提下，随便怎么用都行
EmguCv	GPL v3	http://www.gnu.org/licenses/gpl-3.0.txt	你的产品必须也使用GPL协议，开源且免费
	商业授权	http://www.emgu.com/wiki/files/CommercialLicense.txt	给钱之后可以用于闭源的商业产品
AForge.net	LGPL v3	http://www.gnu.org/licenses/lgpl.html	如果不修改类库源代码，引用该类库的产品可以闭源和（或）收费

以上三种类库都可以用于开发商业产品，但是EmguCv需要付费；因为我只是用来学习和研究，所以这些许可协议对我无所谓。不过鉴于我们身在中国，如果脸皮厚点，去他丫的许可协议。

 

下载

可以很方便的下载到这些类库，下载地址分别为：

类库	下载地址
OpenCv	http://sourceforge.net/projects/opencvlibrary/files/
EmguCv	http://www.emgu.com/wiki/index.php/Download_And_Installation
AForge.net	http://www.aforgenet.com/framework/downloads.html

 

安装

这些类库的安装都比较简单，直接运行安装程序，并点“下一步”即可完成。但是OpenCv在安装完之后还需要一些额外的处理才能在VS2008里面使用，在http://www.opencv.org.cn有一篇名为《VC2008 Express下安装OpenCv 2.0》的文章专门介绍了如何安装OpenCv。

类库	安装难易度	备注
OpenCv	比较容易	VC下使用需要重新编译
EmguCv	容易
AForge.net	容易

相信看这篇文章的人都不会被安装困扰。

 

文档资料 

类库	总体评价	书籍	网站	文档	示例	社区	备注
OpenCv	中等	中英文	中英文	中英文	较多	中文论坛	有中文资料但不完整
EmguCv	少	无	英文	英文	少	英文论坛	论坛人气很差
AForge.net	少	无	英文	英文	少	英文论坛	论坛人气很差

OpenCv有一些中文资料，另外两种的资料全是英文的；不过EmguCv建立在OpenCv的基础上，大部分OpenCv的资料可以用于EmguCv；而AForge.net是原生的.net类库，对GDI+有很多扩展，一些MSDN的资料可以借鉴。如果在查词典的基础上还看不懂英文文档，基本上可以放弃使用这些类库了。

 

易用性

易用性这玩意，主观意志和个人能力对它影响很大，下面是我的看法：

类库	易用性	备注
OpenCv	比较差	OpenCv大多数功能都以C风格函数形式提供，少部分功能以C++类提供。注意：2.0版将更多的功能封装成类了。
EmguCv	比较好	将OpenCv的绝大部分功能都包装成了.net类、结构或者枚举。不过文档不全，还是得对照OpenCv的文档去看才行。
AForge.net	好	纯.net类库，用起来很方便。

最近几年一直用的是C# ，把C和C++忘记得差不多了，况且本来C/C++我就不太熟，所以对OpenCv的看法恐怕有偏见。

 

性能

这些类库能做的事情很多，我选了最基础的部分来进行性能测试，那就是将一幅彩色图像转换成灰度图，然后再将灰度图转换成二值图像。因为图像处理大部分时间都用于内存读写及运算（特别是矩阵运算），所以这两种操作有一定的代表性。

我分别用以下方式实现了图像的灰度化及二值化：（1）C语言调用OpenCv库；（2）C#调用AForge.net库；（3）C#调用EmguCv库；（4）C#中用P/INVOKE的形式调用OpenCv函数；（5）C#调用自己写的灰度和二值化方法。

C语言调用OpenCv

#include " stdafx.h "
#include " cv.h "
#include " cxcore.h "
#include " highgui.h "
#include " winbase.h "

int _tmain( int argc, _TCHAR * argv[])
{
    // 初始化图像
    IplImage * pIplSource = cvLoadImage( " E:\\xrwang\\ImageProcessLearn\\Debug\\wky_tms_2272x1704.jpg " );
    IplImage * pIplGrayscale = cvCreateImage(cvSize(pIplSource -> width,pIplSource -> height),IPL_DEPTH_8U, 1 );
    IplImage * pIplThreshold = cvCreateImage(cvSize(pIplSource -> width,pIplSource -> height),IPL_DEPTH_8U, 1 );
    // 执行灰度化和二值化，并输出所用时间
    LARGE_INTEGER frequency,count1,count2,count3;
    double time1,time2;
    QueryPerformanceFrequency( & frequency);
    for ( int i = 0 ;i < 10 ;i ++ )
    {
        QueryPerformanceCounter( & count1);
        cvCvtColor(pIplSource,pIplGrayscale,CV_BGR2GRAY);
        QueryPerformanceCounter( & count2);
        cvThreshold(pIplGrayscale,pIplThreshold, 128 , 255 ,CV_THRESH_BINARY);
        QueryPerformanceCounter( & count3);
        time1 = ( double ) 1000.0 * (count2.QuadPart - count1.QuadPart) / frequency.QuadPart;
        time2 = ( double ) 1000.0 * (count3.QuadPart - count2.QuadPart) / frequency.QuadPart;
        printf( " 灰度：%g毫秒，二值化：%g毫秒\r\n " ,time1,time2);
    }
    // 显示图像
    cvNamedWindow( " grayscale " , 0 );
    cvNamedWindow( " threshold " , 0 );
    cvResizeWindow( " grayscale " , 600 , 480 );
    cvResizeWindow( " threshold " , 600 , 480 );
    cvShowImage( " grayscale " ,pIplGrayscale);
    cvShowImage( " threshold " ,pIplThreshold);
    cvWaitKey( 0 );
    // 销毁对象
    cvDestroyAllWindows();
    cvReleaseImage( & pIplThreshold);
    cvReleaseImage( & pIplGrayscale);
    cvReleaseImage( & pIplSource);
    return   0 ;
}

 

C#调用各种类库处理图像

using System;
using System.Collections.Generic;
using System.ComponentModel;
using System.Data;
using System.Drawing;
using System.Drawing.Imaging;
using System.Linq;
using System.Text;
using System.Windows.Forms;
using System.Diagnostics;
using System.Runtime.InteropServices;
using AForge.Imaging.Filters;
using Emgu.CV;
using Emgu.CV.Structure;
using Emgu.CV.CvEnum;

namespace ImageProcessLearn
{
    public   partial   class FormMain : Form
    {
        public FormMain()
        {
            InitializeComponent();
        }

        // 窗体加载时
        private   void FormMain_Load( object sender, EventArgs e)
        {
            // 显示原始图像
            pbSource.Image = Image.FromFile( " wky_tms_2272x1704.jpg " );
        }

        // 使用选定的类库处理图像
        private   void btnProcess_Click( object sender, EventArgs e)
        {
            if (rbAForge.Checked)
            {
                ProcessImageWithAforge();
            }
            else   if (rbEmgucv.Checked)
            {
                ProcessImageWithEmgucv();
            }
            else   if (rbOpencv.Checked)
            {
                ProcessImageWithOpencv();
            }
            else   if (rbOwnMethod.Checked)
                ProcessImageWithOwnMethod();
        }

        ///   <summary>
        /// 使用AForge.net处理图像
        ///   </summary>
        private   void ProcessImageWithAforge()
        {
            Stopwatch sw =   new Stopwatch(); // 计时器
            // 灰度
            sw.Start();
            Grayscale grayscaleFilter =   new Grayscale( 0.299 , 0.587 , 0.114 );
            Bitmap bitmapGrayscale = grayscaleFilter.Apply((Bitmap)pbSource.Image);
            sw.Stop();
            double timeGrayscale = sw.Elapsed.TotalMilliseconds;
            if (pbGrayscale.Image !=   null )
            {
                pbGrayscale.Image.Dispose();
                pbGrayscale.Image =   null ;
            }
            pbGrayscale.Image = bitmapGrayscale;
            // 二值化
            sw.Reset();
            sw.Start();
            Threshold thresholdFilter =   new Threshold( 128 );
            Bitmap bitmapThreshold = thresholdFilter.Apply(bitmapGrayscale);
            sw.Stop();
            double timeThreshold = sw.Elapsed.TotalMilliseconds;
            if (pbThreshold.Image !=   null )
            {
                pbThreshold.Image.Dispose();
                pbThreshold.Image =   null ;
            }
            pbThreshold.Image = bitmapThreshold;
            // 输出所用时间
            txtResult.Text +=   string .Format( " 类库：AForge.net，灰度：{0:F05}毫秒，二值化：{1:F05}毫秒\r\n " , timeGrayscale, timeThreshold);
        }

        ///   <summary>
        /// 使用EmguCv处理图像
        ///   </summary>
        private   void ProcessImageWithEmgucv()
        {
            Stopwatch sw =   new Stopwatch(); // 计时器
            // 灰度
            Image < Bgr, Byte > imageSource =   new Image < Bgr, byte > ((Bitmap)pbSource.Image);
            sw.Start();
            Image < Gray, Byte > imageGrayscale = imageSource.Convert < Gray, Byte > ();
            sw.Stop();
            double timeGrayscale = sw.Elapsed.TotalMilliseconds;
            if (pbGrayscale.Image !=   null )
            {
                pbGrayscale.Image.Dispose();
                pbGrayscale.Image =   null ;
            }
            pbGrayscale.Image = imageGrayscale.ToBitmap();
            // 二值化
            sw.Reset();
            sw.Start();
            Image < Gray, Byte > imageThreshold = imageGrayscale.ThresholdBinary( new Gray( 128 ), new Gray( 255 ));
            sw.Stop();
            double timeThreshold = sw.Elapsed.TotalMilliseconds;
            if (pbThreshold.Image !=   null )
            {
                pbThreshold.Image.Dispose();
                pbThreshold.Image =   null ;
            }
            pbThreshold.Image = imageThreshold.ToBitmap();
            // 输出所用时间
            txtResult.Text +=   string .Format( " 类库：EmguCv，灰度：{0:F05}毫秒，二值化：{1:F05}毫秒\r\n " , timeGrayscale, timeThreshold);
        }

        ///   <summary>
        /// 使用Open Cv P/Invoke处理图像
        ///   </summary>
        unsafe   private   void ProcessImageWithOpencv()
        {
            Stopwatch sw =   new Stopwatch(); // 计时器
            // 灰度
            Image < Bgr, Byte > imageSource =   new Image < Bgr, byte > ((Bitmap)pbSource.Image);
            IntPtr ptrSource = Marshal.AllocHGlobal(Marshal.SizeOf( typeof (MIplImage)));
            Marshal.StructureToPtr(imageSource.MIplImage, ptrSource, true );
            sw.Start();
            IntPtr ptrGrayscale = CvInvoke.cvCreateImage(imageSource.Size, IPL_DEPTH.IPL_DEPTH_8U, 1 );
            CvInvoke.cvCvtColor(ptrSource, ptrGrayscale, COLOR_CONVERSION.CV_BGR2GRAY);
            sw.Stop();
            double timeGrayscale = sw.Elapsed.TotalMilliseconds;
            if (pbGrayscale.Image !=   null )
            {
                pbGrayscale.Image.Dispose();
                pbGrayscale.Image =   null ;
            }
            pbGrayscale.Image = ImageConverter.IplImagePointerToBitmap(ptrGrayscale);
            // 二值化
            sw.Reset();
            sw.Start();
            IntPtr ptrThreshold = CvInvoke.cvCreateImage(imageSource.Size, IPL_DEPTH.IPL_DEPTH_8U, 1 );
            CvInvoke.cvThreshold(ptrGrayscale, ptrThreshold, 128d, 255d, THRESH.CV_THRESH_BINARY);
            sw.Stop();
            double timeThreshold = sw.Elapsed.TotalMilliseconds;
            if (pbThreshold.Image !=   null )
            {
                pbThreshold.Image.Dispose();
                pbThreshold.Image =   null ;
            }
            pbThreshold.Image = ImageConverter.IplImagePointerToBitmap(ptrThreshold);
            // 释放资源
            // CvInvoke.cvReleaseImage(ref ptrThreshold);
            // CvInvoke.cvReleaseImage(ref ptrGrayscale);
            Marshal.FreeHGlobal(ptrSource);
            // 输出所用时间
            txtResult.Text +=   string .Format( " 类库：OpenCv P/Invoke，灰度：{0:F05}毫秒，二值化：{1:F05}毫秒\r\n " , timeGrayscale, timeThreshold);
        }

        ///   <summary>
        /// 使用自定义的方法处理图像
        ///   </summary>
        private   void ProcessImageWithOwnMethod()
        {
            Stopwatch sw =   new Stopwatch(); // 计时器
            // 灰度
            sw.Start();
            Bitmap bitmapGrayscale = Grayscale((Bitmap)pbSource.Image);
            sw.Stop();
            double timeGrayscale = sw.Elapsed.TotalMilliseconds;
            if (pbGrayscale.Image !=   null )
            {
                pbGrayscale.Image.Dispose();
                pbGrayscale.Image =   null ;
            }
            pbGrayscale.Image = bitmapGrayscale;
            // 二值化
            sw.Reset();
            sw.Start();
            Bitmap bitmapThreshold = Threshold(bitmapGrayscale, 128 );
            sw.Stop();
            double timeThreshold = sw.Elapsed.TotalMilliseconds;
            if (pbThreshold.Image !=   null )
            {
                pbThreshold.Image.Dispose();
                pbThreshold.Image =   null ;
            }
            pbThreshold.Image = bitmapThreshold;
            // 输出所用时间
            txtResult.Text +=   string .Format( " 类库：自定义方法，灰度：{0:F05}毫秒，二值化：{1:F05}毫秒\r\n " , timeGrayscale, timeThreshold);
        }

        ///   <summary>
        /// 将指定图像转换成灰度图
        ///   </summary>
        ///   <param name="bitmapSource"> 源图像支持3通道或者4通道图像，支持Format24bppRgb、Format32bppRgb和Format32bppArgb这3种像素格式 </param>
        ///   <returns> 返回灰度图，如果转化失败，返回null。 </returns>
        private Bitmap Grayscale(Bitmap bitmapSource)
        {
            Bitmap bitmapGrayscale =   null ;
            if (bitmapSource !=   null   && (bitmapSource.PixelFormat == PixelFormat.Format24bppRgb || bitmapSource.PixelFormat == PixelFormat.Format32bppArgb || bitmapSource.PixelFormat == PixelFormat.Format32bppRgb))
            {
                int width = bitmapSource.Width;
                int height = bitmapSource.Height;
                Rectangle rect =   new Rectangle( 0 , 0 , width, height);
                bitmapGrayscale =   new Bitmap(width, height, PixelFormat.Format8bppIndexed);
                // 设置调色板
                ColorPalette palette = bitmapGrayscale.Palette;
                for ( int i =   0 ; i < palette.Entries.Length; i ++ )
                    palette.Entries[i] = Color.FromArgb( 255 , i, i, i);
                bitmapGrayscale.Palette = palette;
                BitmapData dataSource = bitmapSource.LockBits(rect, ImageLockMode.ReadOnly, bitmapSource.PixelFormat);
                BitmapData dataGrayscale = bitmapGrayscale.LockBits(rect, ImageLockMode.WriteOnly, PixelFormat.Format8bppIndexed);
                byte b, g, r;
                int strideSource = dataSource.Stride;
                int strideGrayscale = dataGrayscale.Stride;
                unsafe
                {
                    byte * ptrSource = ( byte * )dataSource.Scan0.ToPointer();
                    byte * ptr1;
                    byte * ptrGrayscale = ( byte * )dataGrayscale.Scan0.ToPointer();
                    byte * ptr2;
                    if (bitmapSource.PixelFormat == PixelFormat.Format24bppRgb)
                    {
                        for ( int row =   0 ; row < height; row ++ )
                        {
                            ptr1 = ptrSource + strideSource * row;
                            ptr2 = ptrGrayscale + strideGrayscale * row;
                            for ( int col =   0 ; col < width; col ++ )
                            {
                                b =   * ptr1;
                                ptr1 ++ ;
                                g =   * ptr1;
                                ptr1 ++ ;
                                r =   * ptr1;
                                ptr1 ++ ;
                                * ptr2 = ( byte )( 0.114   * b +   0.587   * g +   0.299   * r);
                                ptr2 ++ ;
                            }
                        }
                    }
                    else     // bitmapSource.PixelFormat == PixelFormat.Format32bppArgb || bitmapSource.PixelFormat == PixelFormat.Format32bppRgb
                    {
                        for ( int row =   0 ; row < height; row ++ )
                        {
                            ptr1 = ptrSource + strideGrayscale * row;
                            ptr2 = ptrGrayscale + strideGrayscale * row;
                            for ( int col =   0 ; col < width; col ++ )
                            {
                                b =   * ptr1;
                                ptr1 ++ ;
                                g =   * ptr1;
                                ptr1 ++ ;
                                r =   * ptr1;
                                ptr1 +=   2 ;
                                * ptr2 = ( byte )( 0.114   * b +   0.587   * g +   0.299   * r);
                                ptr2 ++ ;
                            }
                        }
                    }
                }
                bitmapGrayscale.UnlockBits(dataGrayscale);
                bitmapSource.UnlockBits(dataSource);
            }
            return bitmapGrayscale;
        }

        ///   <summary>
        /// 将指定的灰度图像转换成二值图像。如果某个像素的值大于等于阀值，该像素置为白色；否则置为黑色。
        /// 目前支持8bpp和16bpp两种灰度图像的转换，对于8bpp，阀值介于0~255之间；对于16bpp，阀值介于0~65535之间。
        ///   </summary>
        ///   <param name="bitmapGrayscale"> 灰度图像 </param>
        ///   <param name="thresholdValue"> 阀值 </param>
        ///   <returns> 返回转换之后的二值图像；如果转换失败，返回null。 </returns>
        private Bitmap Threshold(Bitmap bitmapGrayscale, int thresholdValue)
        {
            Bitmap bitmapThreshold =   null ;
            if (bitmapGrayscale !=   null )
            {
                int width = bitmapGrayscale.Width;
                int height = bitmapGrayscale.Height;
                Rectangle rect =   new Rectangle( 0 , 0 , width, height);
                PixelFormat pixelFormat = bitmapGrayscale.PixelFormat;
                if (pixelFormat == PixelFormat.Format8bppIndexed)
                {
                    if (thresholdValue >=   0   && thresholdValue <=   255 )
                    {
                        bitmapThreshold = (Bitmap)bitmapGrayscale.Clone();
                        byte white =   255 ;
                        byte black =   0 ;
                        BitmapData data = bitmapThreshold.LockBits(rect, ImageLockMode.ReadWrite, pixelFormat);
                        unsafe
                        {
                            byte * ptrStart = ( byte * )data.Scan0.ToPointer();
                            byte * ptr1;
                            for ( int row =   0 ; row < height; row ++ )
                            {
                                ptr1 = ptrStart + data.Stride * row;
                                for ( int col =   0 ; col < width; col ++ )
                                {
                                    * ptr1 = ( * ptr1 < thresholdValue) ? black : white;
                                    ptr1 ++ ;
                                }
                            }
                        }
                        bitmapThreshold.UnlockBits(data);
                    }
                }
                else   if (pixelFormat == PixelFormat.Format16bppGrayScale)
                {
                    bitmapThreshold = (Bitmap)bitmapGrayscale.Clone();
                    UInt16 white =   65535 ;
                    UInt16 black =   0 ;
                    BitmapData data = bitmapThreshold.LockBits(rect, ImageLockMode.ReadWrite, pixelFormat);
                    unsafe
                    {
                        byte * ptrStart = ( byte * )data.Scan0.ToPointer();
                        UInt16 * ptr1;
                        for ( int row =   0 ; row < height; row ++ )
                        {
                            ptr1 = (UInt16 * )(ptrStart + data.Stride * row);
                            for ( int col =   0 ; col < width; col ++ )
                            {
                                * ptr1 = ( * ptr1 < thresholdValue) ? black : white;
                                ptr1 ++ ;
                            }
                        }
                    }
                    bitmapThreshold.UnlockBits(data);
                }
            }
            return bitmapThreshold;
        }
    }
}

 

     分别用上述5种形式处理10次，记录下运行时间，去掉每种的最大和最小数据，然后计算平均值。结果如下所示（单位是毫秒）：

语言	类库	灰度化	二值化	性能排名
C	OpenCv	16.89721	7.807766	1
C#	Aforge.net	48.9403	25.32473	5
C#	EmguCv	18.86898	13.74628	3
C#	OpenCv(P/Invoke)	18.68938	10.0149	2
C#	自定义处理方法	48.33593	21.46168	4

测试环境如下：CPU-奔腾4 2.4G，内存-512M，操作系统-Windows XP SP2，显卡-nVidia GForce4 64M，进程数-49，线程数-611，句柄数-13004，可用内存101M。

毫无疑问，用C语言调用OpenCv的性能最好，两种纯.net的方式性能最差。

C语言调用OpenCv的处理效果如下所示：

 

C#的处理效果如下：

 

结论

将上面的内容汇总结果如下表所示：

类库	OpenCv	EmguCv	AForge.net
许可协议	BSD	GPL v3或商业授权	LGPL v3
下载	方便	方便	方便
安装	比较容易	容易	容易
文档资料	中等	少	少
易用性	比较差	比较好	好
性能	很好	比较好	不好

综上所述，我的选择是使用EmguCv作为我的图像处理类库，在必要的时候用P/Invoke的形式调用没有被封装的OpenCv函数。