#include "Export.h"
#include "FeatureMatch.h"


// 计算2图像的相似度
// isLesion: true 表示输入图像只裁取了病灶区域，对比两个病灶的相似程度
// false 表示输入图像是整个原始图像，对比两幅图像的相似程度

bool __cdecl CalcuSimilarity(__in unsigned char* imgData1, __in unsigned char* imgData2, bool isLesion,
	                         ImgInfo imgInfo1, ImgInfo imgInfo2, float& outVal)
{
	try
	{
		if (imgData1 == NULL || imgData2 == NULL)
		{
#ifdef _DEBUG
			cout << "图像数据加载失败！" << endl;
#endif
			return false;
		}
		int imgW1 = imgInfo1.imgW;
		int imgH1 = imgInfo1.imgH;
		int imgW2 = imgInfo2.imgW;
		int imgH2 = imgInfo2.imgH;
		int channel = imgInfo1.channel;
		if (imgW1 == 0 || imgW2==0)
		{
			return false;
		}
		if (!(channel == 1 || channel == 3 || channel == 4) )
		{
			return false;
		}
		Mat img1(cv::Size(imgW1, imgH1), CV_8UC(channel), imgData1);
		Mat img2(cv::Size(imgW2, imgH2), CV_8UC(channel), imgData2);
		Mat imgGray1, imgGray2;
		if (channel > 1)
		{
			cv::cvtColor(img1, imgGray1, COLOR_BGR2GRAY);
			cv::cvtColor(img2, imgGray2, COLOR_BGR2GRAY);
		}
		else
		{
			imgGray1 = img1.clone();
			imgGray2 = img2.clone();
		}

		int threWidth;
		if (!isLesion)
		{
			// 对比图像相似程度，需要先做滤波
			float sigma = 3.f;     //高斯核标准差
			int kSize = static_cast<int>(round(2 * sigma) * 2) + 1;
			cv::GaussianBlur(imgGray1, imgGray1, Size(kSize, kSize), sigma, sigma, BORDER_CONSTANT);
			cv::GaussianBlur(imgGray2, imgGray2, Size(kSize, kSize), sigma, sigma, BORDER_CONSTANT);
			threWidth = 80;
		}
		else
		{
			cv::medianBlur(imgGray1, imgGray1, 5);
			cv::medianBlur(imgGray2, imgGray2, 5);
			threWidth = 140;
		}

		if (imgW1 > threWidth)
		{
			int hightNew1 = cvRound((threWidth /(float)imgW1)*imgH1 );
			Size sz(threWidth, hightNew1);
			resize(imgGray1, imgGray1, sz);
		}
		if (imgW2 > threWidth)
		{
			int hightNew2 = cvRound((threWidth /(float)imgW2)*imgH2 );
			Size sz(threWidth, hightNew2);
			resize(imgGray2, imgGray2, sz);
		}

		vector<KeyPoint> keypoints1, keypoints2;
		vector<DMatch> rightMatchs;
		SIFTMatch(imgGray1, imgGray2, keypoints1, keypoints2, rightMatchs);
		int num = (int)rightMatchs.size();   //正确匹配点的个数，以此衡量2张图像的相似度
		if (keypoints1.size() + keypoints2.size() <= 0)
		{
			outVal = 0.f;
		}
		else
		{
			//outVal = 1.0f * num / float(keypoints1.size() + keypoints2.size() - num);  //旧版公式
			outVal = 2.0f * num / float(keypoints1.size() + keypoints2.size());   //相似度计算公式
		}
		if (outVal > 1.0f)
		{
			outVal = 1.0f;
		}
#ifdef _DEBUG
		cout << "正确匹配特征点的数目是：" << num << endl;
		cout << "2幅图的相似度是：" << outVal << endl;
		cout << endl;
#endif
		return true;
	}
	catch (...)
	{
		return false;
	}
}