private void поискПоКадруToolStripMenuItem_Click(object sender, EventArgs e)
{
try
{
Bitmap bmp = (Bitmap)pictureBox1.Image;
Image <Gray, Byte> myImage = new Image <Gray, Byte>(bmp);
DenseHistogram hist = new DenseHistogram(256, new RangeF(0.0f, 255.0f));
hist.Calculate <Byte>(new Image <Gray, byte>[] { myImage }, true, null);
var imList = DBOPS.GetSearchImagesList();
for (int i = 0; i < imList.Count; i++)
{
Bitmap myBmp = (Bitmap)Image.FromFile(System.IO.Directory.GetCurrentDirectory() + imList[i].path);
Image <Gray, Byte> myImage2 = new Image <Gray, Byte>(myBmp);
DenseHistogram hist2 = new DenseHistogram(256, new RangeF(0.0f, 255.0f));
hist2.Calculate <Byte>(new Image <Gray, byte>[] { myImage2 }, true, null);
var a = CvInvoke.CompareHist(hist, hist2, Emgu.CV.CvEnum.HistogramCompMethod.Chisqr);
imList[i].distance = a;
}
var sortedimgs = imList.OrderBy(a => a.distance).ToList();
List <SearchImageListTable> top10img = new List <SearchImageListTable>();
for (int i = 0; i < 10; i++)
{
top10img.Add(sortedimgs[i]);
}
var myForm = new Top10Form(top10img);
//myForm.FormClosed += new FormClosedEventHandler(myForm_FormClosed);
myForm.Show();
}
catch (Exception ex) { }
}
public void Compare(Mat mat1, Mat mat2)
{
CvInvoke.Resize(mat2, mat2, mat1.Size);
CvInvoke.CvtColor(mat1, mat1, ColorConversion.Bgr2Gray);
CvInvoke.CvtColor(mat2, mat2, ColorConversion.Bgr2Gray);
//直方图尺寸设置
//一个灰度值可以设定一个bins，256个灰度值就可以设定256个bins
//对应HSV格式,构建二维直方图
//每个维度的直方图灰度值划分为256块进行统计，也可以使用其他值
int hBins = 256, sBins = 256;
int[] histSize = { hBins };
//H:0~180, S:0~255,V:0~255
//H色调取值范围
float[] hRanges = { 0, 180 };
//S饱和度取值范围
float[] sRanges = { 180 };
float[][] ranges = { new float[] { 0, 40 }, new float[] { 40, 80 }, new float[] { 40, 255 } };
int[] channels = { 0 };//二维直方图
Mat hist1 = new Mat(), hist2 = new Mat();
CvInvoke.CalcHist(mat1, channels, new Mat(), hist1, histSize, sRanges, false);
CvInvoke.CalcHist(mat2, channels, new Mat(), hist2, histSize, sRanges, false);
var result = CvInvoke.CompareHist(hist1, hist2, HistogramCompMethod.Correl);
Console.WriteLine($"result:{result}");
}
private bool CompareHistogram(string imgOne, string imgTwo)
{
// based on the outdated code from: http://www.emgu.com/forum/viewtopic.php?t=2956
var imageOne = new Image <Rgb, byte>(imgOne);
var imageTwo = new Image <Rgb, byte>(imgTwo);
// separate image one by channel
DenseHistogram imageOneBlueHist = new DenseHistogram(64, new RangeF(0, 64));
DenseHistogram imageOneGreenHist = new DenseHistogram(64, new RangeF(0, 64));
DenseHistogram imageOneRedHist = new DenseHistogram(64, new RangeF(0, 64));
Image <Gray, Byte> imageOneBlue = imageOne[2];
Image <Gray, Byte> imageOneGreen = imageOne[1];
Image <Gray, Byte> imageOneRed = imageOne[0];
// calculate histogram of image one
imageOneBlueHist.Calculate(new Image <Gray, Byte>[] { imageOneBlue }, true, null);
imageOneGreenHist.Calculate(new Image <Gray, Byte>[] { imageOneGreen }, true, null);
imageOneRedHist.Calculate(new Image <Gray, Byte>[] { imageOneRed }, true, null);
// separate image two by channels
DenseHistogram imageTwoBlueHist = new DenseHistogram(64, new RangeF(0, 64));
DenseHistogram imageTwoGreenHist = new DenseHistogram(64, new RangeF(0, 64));
DenseHistogram imageTwoRedHist = new DenseHistogram(64, new RangeF(0, 64));
Image <Gray, Byte> imageTwoBlue = imageTwo[2];
Image <Gray, Byte> imageTwoGreen = imageTwo[1];
Image <Gray, Byte> imageTwoRed = imageTwo[0];
// calculate histogram of image two
imageTwoBlueHist.Calculate(new Image <Gray, Byte>[] { imageTwoBlue }, true, null);
imageTwoGreenHist.Calculate(new Image <Gray, Byte>[] { imageTwoGreen }, true, null);
imageTwoRedHist.Calculate(new Image <Gray, Byte>[] { imageTwoRed }, true, null);
// comparing histograms
double blueResult
= CvInvoke.CompareHist(
imageTwoBlueHist, imageOneBlueHist, Emgu.CV.CvEnum.HistogramCompMethod.Correl);
double greenResult
= CvInvoke.CompareHist(
imageTwoGreenHist, imageOneGreenHist, Emgu.CV.CvEnum.HistogramCompMethod.Correl);
double redResult
= CvInvoke.CompareHist(
imageTwoRedHist, imageOneRedHist, Emgu.CV.CvEnum.HistogramCompMethod.Correl);
// judge the comparison result
if ((blueResult * greenResult * redResult) >= HISTOGRAM_SIMILARITY)
{
return(true);
}
else
{
return(false);
}
}
public static double MatchHistograms(List <Mat> hist1, List <Mat> hist2)
{
double distance = 0;
for (int i = 0; i < hist1.Count; i++)
{
double tmp = CvInvoke.CompareHist(hist1[i], hist2[i], HistogramCompMethod.ChisqrAlt);
distance += tmp;
}
return(distance);
}
private double CheckSimilarity(ref Mat refStateImg)
{
Mat tempImg = CaptureMatWindow();
double compareRatio = CvInvoke.CompareHist(refStateImg, tempImg, Emgu.CV.CvEnum.HistogramCompMethod.Chisqr);
Application.Current.Dispatcher.Invoke(DispatcherPriority.Normal, new Action(delegate
{
((MainWindow)Application.Current.MainWindow).img_reference.Source = BitmapSourceExtension.ToBitmapSource(tempImg);
((MainWindow)Application.Current.MainWindow).btn_start.Content = Convert.ToString(Convert.ToInt32(compareRatio));
}));
return(compareRatio);
}
private void button5_Click(object sender, EventArgs e)
{
//在比较直方图时，最佳操作是在HSV空间中操作，所以需要将BGR空间转换为HSV空间
Mat srcHsvImage = new Mat(CvInvoke.cvGetSize(src1), DepthType.Cv8U, 3);
Mat compareHsvImage = new Mat(CvInvoke.cvGetSize(src2), DepthType.Cv8U, 3);
CvInvoke.CvtColor(src1, srcHsvImage, ColorConversion.Bgr2Hsv);
CvInvoke.CvtColor(src2, compareHsvImage, ColorConversion.Bgr2Hsv);
//采用H-S直方图进行处理
//首先得配置直方图的参数
Mat srcHist = new Mat(CvInvoke.cvGetSize(src1), DepthType.Cv8U, 3);
Mat compHist = new Mat(CvInvoke.cvGetSize(src2), DepthType.Cv8U, 3);
//H、S通道
int[] channels = new int[2] {
0, 1
};
int[] histSize = new int[2] {
30, 32
};
float[] Ranges = new float[2] {
0, 180
};
//进行原图直方图的计算
CvInvoke.CalcHist(srcHsvImage, channels, null, srcHist, histSize, Ranges, true);
//对需要比较的图进行直方图的计算
CvInvoke.CalcHist(compareHsvImage, channels, null, compHist, histSize, Ranges, true);
//注意：这里需要对两个直方图进行归一化操作
CvInvoke.Normalize(srcHist, srcHist, 1, 0, NormType.MinMax);
CvInvoke.Normalize(compHist, compHist, 1, 0, NormType.MinMax);
//对得到的直方图对比
//相关：CV_COMP_CORREL
//卡方：CV_COMP_CHISQR
//直方图相交：CV_COMP_INTERSECT
//Bhattacharyya距离：CV_COMP_BHATTACHARYYA
double g_dCompareRecult = CvInvoke.CompareHist(srcHist, compHist, his[comboBox1.SelectedIndex]);
richTextBox1.Text = "方法 " + comboBox1.SelectedIndex + "：两幅图像比较的结果为：" + g_dCompareRecult + "\r\n";
}
public static double CompareHistograms(Mat img, Mat img2)
{
using (Mat hist = new Mat())
using (Mat hist2 = new Mat())
using (VectorOfMat vm = new VectorOfMat())
using (VectorOfMat vm2 = new VectorOfMat()) {
vm.Push(img);
vm2.Push(img2);
var channels = new int[] { 0 };
var histSize = new int[] { 256 };
var ranges = new float[] { 0, 256, };
CvInvoke.CalcHist(vm, channels, null, hist, histSize, ranges, false);
CvInvoke.CalcHist(vm2, channels, null, hist2, histSize, ranges, false);
//CvInvoke.Normalize(hist, hist, 0, 255, NormType.MinMax);
//CvInvoke.Normalize(hist2, hist2, 0, 255, NormType.MinMax);
//double res = CvInvoke.CompareHist(hist, hist2, HistogramCompMethod.Bhattacharyya);
//Debug.Log("Cards in Stock: " + (res > 0.5));
return(CvInvoke.CompareHist(hist, hist2, HistogramCompMethod.Correl));
}
}
public double calculateDifference()//简化代码量，直接默认输入图像都是BGR三通道彩色图像
{
DenseHistogram[] histogramOfBaseImage = new DenseHistogram[3];
DenseHistogram[] histogramOfInputImage = new DenseHistogram[3];
for (int i = 0; i < 3; i++)
{
histogramOfBaseImage[i] = new DenseHistogram(nBins, new RangeF(0f, 255.1f));
histogramOfBaseImage[i].Calculate <byte>(new Image <Gray, byte>[] { baseImg.ToImage <Bgr, byte>().Split()[i] }, false, null);
histogramOfInputImage[i] = new DenseHistogram(nBins, new RangeF(0f, 255.1f));
histogramOfInputImage[i].Calculate <byte>(new Image <Gray, byte>[] { inputImg.ToImage <Bgr, byte>().Split()[i] }, false, null);
}
Image <Bgr, float> histOfBaseImg = new Image <Bgr, float>(nBins, 1);
Image <Bgr, float> histOfInputImg = new Image <Bgr, float>(nBins, 1);
for (int i = 0; i < 3; i++)
{
for (int j = 0; j < nBins; j++)
{
histOfBaseImg.Data[0, j, i] = (float)histogramOfBaseImage[i].GetBinValues()[j];
histOfInputImg.Data[0, j, i] = (float)histogramOfInputImage[i].GetBinValues()[j];
}
}
return(CvInvoke.CompareHist(histOfBaseImg, histOfInputImg, HistogramCompMethod.Bhattacharyya));//也可以不用巴氏系数，采用其他比较方法
}
public void Detection()
{
try
{
check = comp1;
for (int i = 0; i < dt.Rows.Count; i++)
{
capturedImg = new Image <Gray, byte>(faceDetected);
DBImg = new Image <Gray, byte>((dt.Rows[i].ItemArray[0]).ToString());
hist1 = new DenseHistogram(256, new RangeF(0.0f, 255.0f));
hist2 = new DenseHistogram(256, new RangeF(0.0f, 255.0f));
hist1.Calculate(new Image <Gray, byte>[] { capturedImg }, false, null);
hist2.Calculate(new Image <Gray, byte>[] { DBImg }, false, null);
mat1 = new Mat();
hist1.CopyTo(mat1);
mat2 = new Mat();
hist2.CopyTo(mat2);
comp1 = CvInvoke.CompareHist(mat1, mat2, Emgu.CV.CvEnum.HistogramCompMethod.Bhattacharyya);
if (comp1 > check)
{
check = comp1;
index = i;
}
}
MessageBox.Show(check + " " + index.ToString());
}
catch (Exception ee)
{
MessageBox.Show(ee.Message);
}
}
public double CompareTo(PageMetric other)
{
return(CvInvoke.CompareHist(this.histogram, other.histogram, HistogramCompMethod.Chisqr));
}
public string Predict(Matrix <float> data, out Dictionary <string, float> probabilities, int[] best = null)
{
float response = -1;
probabilities = new Dictionary <string, float>();
if (type == ClassifierType.SVM)
{
if (svm == null)
{
if (nameForID.Count > 0)
{
return(nameForID[0]);
}
else
{
return("null");
}
}
if (nameForID.Count == 1)
{
response = 0;
}
else
{
int numFeatures = Math.Max(data.Cols, 2);
double[] input = new double[numFeatures];
for (int i = 0; i < numFeatures; i++)
{
input[i] = data[0, Math.Min(data.Cols - 1, i)];
}
double[] responses = new double[nameForID.Count];
response = svm.Compute(input, out responses);
for (int i = 0; i < responses.Length; i++)
{
probabilities.Add(nameForID[i], (float)responses[i]);
}
}
}
else if (type == ClassifierType.NeuralNet)
{
double[] input = new double[data.Cols];
for (int i = 0; i < data.Cols; i++)
{
input[i] = data[0, i];
}
double[] fusion = network.Compute(input);
int prediction; fusion.Max(out prediction);
response = prediction;
}
else if (type == ClassifierType.KMeans)
{
double[] dataRow = new double[data.Cols];
for (int i = 0; i < dataRow.Length; i++)
{
dataRow[i] = data[0, i];
}
if (kmeans == null)
{
return("null");
}
List <Tuple <double, int> > clusterDistances = new List <Tuple <double, int> >();
for (int i = 0; i < kmeans.Clusters.Count; i++)
{
double dist = Utils.DistL2Sqr(kmeans.Clusters[i].Mean, dataRow);
clusterDistances.Add(new Tuple <double, int>(dist, i));
}
clusterDistances.Sort((Tuple <double, int> a, Tuple <double, int> b) => { return(a.Item1.CompareTo(b.Item1)); });
int mostInstances = 0;
int totalCount = 0;
foreach (int classLabel in clusterClasses[clusterDistances[0].Item2].Keys)
{
int instances = clusterClasses[clusterDistances[0].Item2][classLabel];
totalCount += instances;
if (instances > mostInstances)
{
mostInstances = instances;
response = classLabel;
}
}
}
else
{
if (trainData == null)
{
return("error: no training data");
}
List <Tuple <double, int, float> > results = new List <Tuple <double, int, float> >();
for (int i = 0; i < trainData.Rows; i++)
{
double dist = CvInvoke.CompareHist(trainData.GetRow(i), data, HistogramCompMethod.Chisqr);
results.Add(new Tuple <double, int, float>(dist, i, trainLabels[i, 0]));
}
results.Sort((Tuple <double, int, float> a, Tuple <double, int, float> b) => { return(a.Item1.CompareTo(b.Item1)); });
response = results[0].Item3;
if (best != null)
{
int n = Math.Min(results.Count, best.Length);
for (int i = 0; i < n; i++)
{
best[i] = results[i].Item2;
}
}
}
int classID = (int)Math.Round(response);
string className = (classID >= 0 && classID < nameForID.Count) ? nameForID[classID] : "no_match";
return(className);
}
public void Detection()
{
try
{
check = comp1;
for (int i = 0; i < dt.Rows.Count; i++)
{
capturedImg = new Image <Gray, byte>(pbDetectedFace.Image.Bitmap);
DBImg = new Image <Gray, byte>((dt.Rows[i].ItemArray[0]).ToString());
hist1 = new DenseHistogram(256, new RangeF(0.0f, 255.0f));
hist2 = new DenseHistogram(256, new RangeF(0.0f, 255.0f));
hist1.Calculate(new Image <Gray, byte>[] { capturedImg }, false, null);
hist2.Calculate(new Image <Gray, byte>[] { DBImg }, false, null);
mat1 = new Mat();
hist1.CopyTo(mat1);
mat2 = new Mat();
hist2.CopyTo(mat2);
//float[] histFloat = new float[256];
//hist1.CopyTo(histFloat);
//float[] hist2Float = new float[256];
//hist2.CopyTo(hist2Float);
//double count1 = 0, count2 = 0;
//for (int j = 0; j < 256; j++)
//{
// count1 += histFloat[j];
// count2 += hist2Float[j];
//}
//unsafe
//{
// fixed (char* ch = (dt.Rows[i].ItemArray[0]).ToString().ToCharArray())
// {
// fixed (char* ch2 = (dt.Rows[i].ItemArray[0]).ToString().ToCharArray())
// {
// OpenCVWrapper.OpencvWrapperClass obj = new OpencvWrapperClass();
// sbyte* pic1 = (sbyte*)ch2;
// sbyte* pic2 = (sbyte*)ch;
// comp1=obj.CompareHistogram(pic1, pic2);
// }
// }
//}
comp1 = CvInvoke.CompareHist(mat1, mat2, Emgu.CV.CvEnum.HistogramCompMethod.Correl);
if (comp1 > check && comp1 < 0.40 && comp1 > 0.221)
{
check = comp1;
index = i;
}
}
MessageBox.Show(check + " " + index.ToString());
}
catch (IndexOutOfRangeException index)
{
MessageBox.Show(index.Message);
}
catch (Exception ee)
{
MessageBox.Show(ee.Message);
}
}
private void compareToolStripMenuItem_Click(object sender, EventArgs e)
{
try
{
if (pictureBox1.Image == null)
{
return;
}
var img = new Bitmap(pictureBox1.Image)
.ToImage <Gray, byte>();
Image <Gray, byte> img1 = null;
OpenFileDialog dialog = new OpenFileDialog();
if (dialog.ShowDialog() == DialogResult.OK)
{
img1 = new Image <Gray, byte>(dialog.FileName);
}
Mat hist = new Mat();
Mat hist1 = new Mat();
float[] ranges = new float[] { 0, 256 };
int[] channel = { 0 };
int[] histSize = { 256 };
VectorOfMat ms = new VectorOfMat();
ms.Push(img);
VectorOfMat ms1 = new VectorOfMat();
ms1.Push(img1);
CvInvoke.CalcHist(ms, channel, null, hist, histSize, ranges, false);
CvInvoke.CalcHist(ms1, channel, null, hist1, histSize, ranges, false);
CvInvoke.Normalize(hist, hist);
CvInvoke.Normalize(hist1, hist1);
HistogramViewer viewer = new HistogramViewer();
viewer.Text = "Image Histogram";
viewer.ShowIcon = false;
viewer.HistogramCtrl.AddHistogram("Image1 Histogram", Color.Blue, hist, 256, ranges);
viewer.HistogramCtrl.Refresh();
viewer.Show();
HistogramViewer viewer1 = new HistogramViewer();
viewer1.Text = "Image Histogram";
viewer1.ShowIcon = false;
viewer1.HistogramCtrl.AddHistogram("Image2 Histogram", Color.Blue, hist1, 256, ranges);
viewer1.HistogramCtrl.Refresh();
viewer1.Show();
var result1 = CvInvoke.CompareHist(hist, hist, Emgu.CV.CvEnum.HistogramCompMethod.Correl);
var result2 = CvInvoke.CompareHist(hist1, hist1, Emgu.CV.CvEnum.HistogramCompMethod.Correl);
var result3 = CvInvoke.CompareHist(hist, hist1, Emgu.CV.CvEnum.HistogramCompMethod.Correl);
lblBGR.Text = "Hist vs Hist = " + result1.ToString() + "\n" +
"Hist1 vs Hist1 = " + result2.ToString() + "\n" +
"Hist vs Hist1 = " + result3.ToString() + "\n";
//pictureBox1.Image = CreateGraph(hist).GetImage();
}
catch (Exception ex)
{
MessageBox.Show(ex.Message);
}
}
