版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/qinghuan1/article/details/78971246
连通域在图像处理领域用的还是比较多的，在网上搜索连通域提取的算法和代码，比较多的是用C++写的，浏览次数比较多的是这篇博客，但由于一直是用C#编写代码，所以仿照那篇博客中的C++程序，将其改为了C#程序，同时在此基础上添加了质心计算和标记函数。
构造了两个结构体，EqualRun用于记录两个等价run，centroid用于存储与质心计算相关的参数。
public struct EqualRun
{
public int firstRun;
public int secondRun;
};
public struct Centroid
{
public int sumx;
public int sumy;
public int area;
};此三个函数：fillRunVectors、firstpass、replaceSameLabel是仿照博客写的。函数centroidCalculate用于计算连通域的质心。函数drawCaliResult用于在图中标记出质心位置。
//fillRunVectors函数完成所有团的查找与记录
public void fillRunVectors(Bitmap bwBitmap,ref int NumberOfRuns, ref List<int> stRun,
ref List<int> enRun, ref List<int> rowRun, out List<Point> stPoint, out List<Point> endPoint)
{
stPoint = new List<Point>();
endPoint = new List<Point>();
BitmapData bwBmpData = bwBitmap.LockBits(new Rectangle(0, 0, bwBitmap.Width, bwBitmap.Height), ImageLockMode.ReadWrite, bwBitmap.PixelFormat);
byte* sorPtr = (byte*)bwBmpData.Scan0;
for (int i = 0; i<bwBitmap.Height; i++)//行
{
if (sorPtr[i* bwBmpData.Stride] == 255)//每行的第一个元素
{
NumberOfRuns++;
stRun.Add(0);
rowRun.Add(i);
stPoint.Add(new Point(0, i));
}
for (int j = 1; j<bwBitmap.Width; j++)//列
{
if (sorPtr[i * bwBmpData.Stride + j - 1] == 0 && sorPtr[i * bwBmpData.Stride + j] == 255)//新增一个团
{
NumberOfRuns++;
stRun.Add(j);
rowRun.Add(i);
stPoint.Add(new Point(j, i));
}
else if (sorPtr[i * bwBmpData.Stride + j - 1] == 255 && sorPtr[i * bwBmpData.Stride + j] == 0)//结束一个团
{
enRun.Add(j - 1);
endPoint.Add(new Point(j - 1, i));
}
}
if (sorPtr[i * bwBmpData.Stride + bwBmpData.Width - 1]==255)//每行的最后一个元素
{
enRun.Add(bwBitmap.Width - 1);
endPoint.Add(new Point(bwBitmap.Width - 1, i));
}
}
bwBitmap.UnlockBits(bwBmpData);
}
//firstPass函数完成团的标记与等价对列表的生成 offset=1,8邻域，offset=0，四邻域
public void firstPass(ref List<int> stRun, ref List<int> enRun, ref List<int> rowRun, ref int NumberOfRuns,
ref List<int> runLabels, ref List<EqualRun> equivalences, int offset)
{
//runLabels = new List<int>(NumberOfRuns);//可能会出问题
for (int i = 0; i < NumberOfRuns; i++)
{
runLabels.Add(0);
}
int idxLabel = 1;
int curRowIdx = 0;
int firstRunOnCur = 0;
int firstRunOnPre = 0;
int lastRunOnPre = -1;
for (int i = 0; i < NumberOfRuns; i++)
{
// 如果是该行的第一个run，则更新上一行第一个run第最后一个run的序号
if (rowRun[i] != curRowIdx)
{
curRowIdx = rowRun[i]; // 更新行的序号
firstRunOnPre = firstRunOnCur;
lastRunOnPre = i - 1;
firstRunOnCur = i;
}
// 遍历上一行的所有run，判断是否与当前run有重合的区域
for (int j = firstRunOnPre; j <= lastRunOnPre; j++)
{
// 区域重合 且 处于相邻的两行
if (stRun[i] <= enRun[j] + offset && enRun[i] >= stRun[j] - offset && rowRun[i] == rowRun[j] + 1)
{
if (runLabels[i] == 0) // 没有被标号过
runLabels[i] = runLabels[j];
else if (runLabels[i] != runLabels[j])// 已经被标号
{
EqualRun equalRun;
equalRun.firstRun = runLabels[i];
equalRun.secondRun = runLabels[j];
equivalences.Add(equalRun); // 保存等价对
}
}
}
if (runLabels[i] == 0) // 没有与前一列的任何run重合
{
runLabels[i] = idxLabel++;
}
}
}
//每个等价表用一个vector<int>来保存，等价对列表保存在map<pair<int,int>>里
public void replaceSameLabel(ref List<int> runLabels, ref List<EqualRun> equivalence, out List<List<int>> equaList)
{
int maxLabel = runLabels.Max();//找到最大的标号
bool[,] eqTab = new bool[maxLabel + 1, maxLabel + 1];
for(int i = 0; i < maxLabel; i++)
{
for(int j = 0; j < maxLabel; j++)
{
eqTab[i, j] = false;
}
}
foreach(var item in equivalence)
{
eqTab[item.firstRun, item.secondRun] = true;
eqTab[item.secondRun, item.firstRun] = true;
}
List<int> labelFlag = new List<int>();
for (int i = 0; i < maxLabel + 1; i++)
{
labelFlag.Add(0);
}
equaList = new List<List<int>>();//等价链表
/*
遍历所有的团标号，来看每个标号应该加入哪个连通域，并赋连通域标号（labelFlag）。
若当前的标号不属于任何一个连通域，则新给它一个连通域标号（labelFlag），并把它加入当前的连通域（tempList.push_back()），
遍历当前连通域内的所有标号，将包含其中标号的所有等价对都加入到当前连通域内。
*/
int labelFlagNum = 1;
for (int i = 1; i <= maxLabel; i++)
{
if (labelFlag[i] != 0)
{
continue;
}
labelFlag[i] = labelFlagNum;
List<int> tempList = new List<int>();
tempList.Add(i);
for (int j = 0; j < tempList.Count; j++)
{
for (int k = 0; k != maxLabel; k++)
{
if (eqTab[tempList[j], k] && labelFlag[k]==0) //说明标号k+1与j连通，并且k+1没有被表示过，不属于任何连通域
{
tempList.Add(k);
labelFlag[k] = labelFlagNum;
}
}
}
equaList.Add(tempList);
labelFlagNum++;
}
for (int i = 0; i != runLabels.Count; i++)
{
runLabels[i] = labelFlag[runLabels[i]];
}
}public void centroidCalculate(ref List<Point> stPoint, ref List<Point> endPoint, ref List<int> runLabels, out List<Point> centroidPointList)
{
int maxLabel = runLabels.Max();//找到最大的标号
List<Centroid> centroidList = new List<Centroid>();//保存与质心计算有关的参数，总个数为连通域个数
for(int i = 0; i < maxLabel; i++)
{
Centroid tempCentroid;
tempCentroid.sumx = 0;
tempCentroid.sumy = 0;
tempCentroid.area = 0;
centroidList.Add(tempCentroid);
}
for(int i = 0; i< runLabels.Count; i++)
{
int index = runLabels[i];
int tempSumx = (stPoint[i].X + endPoint[i].X) * (endPoint[i].X - stPoint[i].X + 1) / 2;
int tempSumy = (endPoint[i].X - stPoint[i].X + 1) * stPoint[i].Y;
int tempArea = endPoint[i].X - stPoint[i].X + 1;
int newSumx = centroidList[index - 1].sumx + tempSumx;
int newSumy = centroidList[index - 1].sumy + tempSumy;
int newArea = centroidList[index - 1].area + tempArea;
Centroid newCentroid;
newCentroid.sumx = newSumx; newCentroid.sumy = newSumy; newCentroid.area = newArea;
centroidList[index - 1] = newCentroid;
}
centroidPointList = new List<Point>();
for(int i = 0; i < maxLabel; i++)
{
Point tempPoint = new Point(centroidList[i].sumx / centroidList[i].area, centroidList[i].sumy / centroidList[i].area);
centroidPointList.Add(tempPoint);
}
} /// <summary>
/// 画出连通域的质心
/// </summary>
/// <param name="oriBitmap">二值图像</param>
/// <param name="centroidPointList">质心点集</param>
/// <returns></returns>
public Bitmap drawCaliResult(Bitmap oriBitmap, List<Point> centroidPointList)
{
Bitmap result = (Bitmap)oriBitmap.Clone();
Graphics g;
Color[] color = new Color[9];
color[0] = Color.FromArgb(255, 0, 0);
color[1] = Color.FromArgb(0, 255, 0);
color[2] = Color.FromArgb(0, 0, 255);
color[3] = Color.FromArgb(255, 255, 0);
color[4] = Color.FromArgb(0, 255, 255);
color[5] = Color.FromArgb(255, 0, 255);
color[6] = Color.FromArgb(125, 0, 125);
color[7] = Color.FromArgb(125, 125, 0);
color[8] = Color.FromArgb(0, 125, 125);
using (Image img = result)
{
//如果原图片是索引像素格式之列的，则需要转换格式
if (IsPixelFormatIndexed(img.PixelFormat))
{
result = new Bitmap(img.Width, img.Height, PixelFormat.Format32bppArgb);
using (g = Graphics.FromImage(result))
{
g.InterpolationMode = System.Drawing.Drawing2D.InterpolationMode.HighQualityBicubic;
g.SmoothingMode = System.Drawing.Drawing2D.SmoothingMode.HighQuality;
g.CompositingQuality = System.Drawing.Drawing2D.CompositingQuality.HighQuality;
g.DrawImage(img, 0, 0);
if (centroidPointList != null || centroidPointList.Count > 0)
{
for (int i = 0; i < centroidPointList.Count; ++i)
{
Brush brush = new SolidBrush(color[0]);
g.FillEllipse(brush, centroidPointList[i].X, centroidPointList[i].Y, 3, 3);
}
}
g.Dispose();
}
}
else //否则直接操作
{
g = Graphics.FromImage(result);
if (centroidPointList != null || centroidPointList.Count > 0)
{
for (int i = 0; i < centroidPointList.Count; ++i)
{
Brush brush = new SolidBrush(color[0]);
g.FillEllipse(brush, centroidPointList[i].X, centroidPointList[i].Y, 3, 3);
}
}
g.Dispose();
}
}
return result;
}
/// <summary>
/// 会产生graphics异常的PixelFormat
/// </summary>
private static PixelFormat[] indexedPixelFormats = { PixelFormat.Undefined, PixelFormat.DontCare,PixelFormat.Format16bppArgb1555,
PixelFormat.Format1bppIndexed, PixelFormat.Format4bppIndexed, PixelFormat.Format8bppIndexed};
/// <summary>
/// 判断图片的PixelFormat 是否在引发异常的 PixelFormat 之中
/// </summary>
/// <param name="imgPixelFormat">原图片的PixelFormat</param>
/// <returns></returns>
private static bool IsPixelFormatIndexed(PixelFormat imgPixelFormat)
{
foreach (PixelFormat pf in indexedPixelFormats)
{
if (pf.Equals(imgPixelFormat)) return true;
}
return false;
}
具体调用方法如下，其中参数openImage应为二值图像。
int numberOfRuns = 0;
List<int> stRun = new List<int>(); List<int> enRun = new List<int>(); List<int> rowRun = new List<int>();
List<Point> stPoint; List<Point> endPoint; List<List<int>> equaList; List<Point> centroidPointList;
List<int> runLabels = new List<int>(); List<EqualRun> equivalences = new List<EqualRun>();
curImageProcess.fillRunVectors(openImage, ref numberOfRuns, ref stRun, ref enRun, ref rowRun, out stPoint, out endPoint);
curImageProcess.firstPass(ref stRun, ref enRun, ref rowRun, ref numberOfRuns, ref runLabels, ref equivalences, 1);
curImageProcess.replaceSameLabel(ref runLabels, ref equivalences, out equaList);
curImageProcess.centroidCalculate(ref stPoint, ref endPoint, ref runLabels, out centroidPointList);
Bitmap resBitmap = curImageProcess.drawCaliResult(openImage, centroidPointList);
