基于开源算法实现图片比对进行图片全图和局部 比对

  public IEnumerable<TFeature>[] Transform(Bitmap[] input)
        {
            return Transform(input, new IList<TFeature>[input.Length]);
        }

        public IEnumerable<TFeature>[] Transform(Bitmap[] input, IEnumerable<TFeature>[] result)
        {
            for (int i = 0; i < input.Length; i++)
                result[i] = Transform(input[i]);
            return result;
        }

 private void btnFreak_Click(object sender, EventArgs e)
        {
            if (string.IsNullOrWhiteSpace(label1.Text) || string.IsNullOrWhiteSpace(label2.Text))
            {

                MessageBox.Show("请先上传比对图片！");

                return;
            }

          
            Concatenate concatenate1 = new Concatenate(img1);
            pictureBox.Image = concatenate1.Apply(img2);

         
            FastRetinaKeypointDetector freak = new FastRetinaKeypointDetector();
            //获取两张图片的采样点
            keyPoints1 = freak.Transform(img1);
            keyPoints2 = freak.Transform(img2);

       
            Bitmap img1mark = new PointsMarker(keyPoints1.Select(x => (IFeaturePoint)x).ToList()).Apply(img1);
            Bitmap img2mark = new PointsMarker(keyPoints2.Select(x => (IFeaturePoint)x).ToList()).Apply(img2);

           
            Concatenate concatenate2 = new Concatenate(img1mark);
            pictureBox.Image = concatenate2.Apply(img2mark);
        }

    private void btnCorrelation_Click(object sender, EventArgs e)
        {
            if (keyPoints1 == null)
            {
                MessageBox.Show("Please, click FREAK button first! :-)");
                return;
            }



            Thread t1 = new Thread(new ThreadStart(() =>
            {
                this.Invoke(new Action(() =>
                {
                    lb_result.Text = "检测中......";
                }));

                var matcher = new KNearestNeighborMatching<byte[]>(5, new Hamming());
                IntPoint[][] matches = matcher.Match(keyPoints1, keyPoints2);//返回的结果值一般都是相同的，数组中
                string result1 = (((decimal)matches[0].Count() / (decimal)keyPoints1.Count()) * 100).ToString("#0.00");
                string result2 = (((decimal)matches[1].Count() / (decimal)keyPoints2.Count()) * 100).ToString("#0.00");
                string tempString = string.Format("原图采样点：{0},对比图采样点：{1},匹配个数点：{2},{3},匹配比例：{4}%,{5}%",
                    keyPoints1.Count(), keyPoints2.Count(), matches[0].Count(), matches[1].Count(), result1, result2);

                this.Invoke(new Action(() =>
                {
                    lb_result.Text = tempString;
                    lb_result.BackColor = Color.Azure;
                
                    correlationPoints1 = matches[0];
                    correlationPoints2 = matches[1];

                
                    Concatenate concat = new Concatenate(img1);
                    Bitmap img3 = concat.Apply(img2);

                   
                    PairsMarker pairs = new PairsMarker(
                        correlationPoints1, // Add image1's width to the X points to show the markings correctly
                        correlationPoints2.Apply(p => new IntPoint(p.X + img1.Width, p.Y)));

                    pictureBox.Image = pairs.Apply(img3);

                }));

            }));

            t1.Start();

        }

  private IntPoint[][] match(IFeaturePoint<T>[] points1, IFeaturePoint<T>[] points2)
        {
            if (points1.Length == 0 || points2.Length == 0)
                throw new ArgumentException("Insufficient number of points to produce a matching.");


            bool swap = false;



            if (points2.Length > points1.Length)
            {
                var aux = points1;
                points1 = points2;
                points2 = aux;
                swap = true;
            }


        
            T[] features1 = new T[points1.Length];
            for (int i = 0; i < features1.Length; i++)
                features1[i] = points1[i].Descriptor;

            T[] features2 = new T[points2.Length];
            for (int i = 0; i < features2.Length; i++)
                features2[i] = points2[i].Descriptor;

            var knn = CreateNeighbors(features1);

            double[] scores = new double[features2.Length];
            int[] labels = new int[features2.Length];
            knn.Score(features2, ref labels, result: scores);

            int[] bestMatch = new int[points1.Length];
            double[] bestScore = new double[points1.Length];
            for (int i = 0; i < bestScore.Length; i++)
                bestScore[i] = Double.PositiveInfinity;

  
            for (int j = 0; j < labels.Length; j++)
            {
                int i = labels[j];

                if (scores[j] > Threshold)
                {
                    if (scores[j] < bestScore[i])
                    {
                        bestScore[i] = scores[j];
                        bestMatch[i] = j;
                    }
                }
            }


            var p1 = new List<IntPoint>(bestScore.Length);
            var p2 = new List<IntPoint>(bestScore.Length);

           
            for (int i = 0; i < bestScore.Length; i++)
            {
                IFeaturePoint<T> pi = points1[i];

                if (bestScore[i] != Double.PositiveInfinity)
                {
                    int j = bestMatch[i];
                    IFeaturePoint<T> pj = points2[j];
                    p1.Add(new IntPoint((int)pi.X, (int)pi.Y));
                    p2.Add(new IntPoint((int)pj.X, (int)pj.Y));
                }
            }

            IntPoint[] m1 = p1.ToArray();
            IntPoint[] m2 = p2.ToArray();

          

            if (swap)
                return new IntPoint[][] { m2, m1 };
            return new IntPoint[][] { m1, m2 };
        }