实验1 图像的灰度变换
实验要求
- 利用OpenCV读取图像
具体内容：用打开OpenCV打开图像，并在窗口中显示。 - 灰度图像二值化处理
具体内容：设置并调整阈值对图像进行二值化处理。 - 灰度图像的对数变换
具体内容：设置并调整r值对图像进行对数变换。 - 灰度图像的伽马变换
具体内容：设置并调整γ值对图像进行伽马变换。 - 彩色图像的补色变换
具体内容：对彩色图像进行补色变换。
实验完成情况
包括完成的实验内容及每个实验的完成程度
完成了8项内容：
- 利用OpenCV读取图像。
int ShowImg() { namedWindow("Original - 原始图像", WINDOW_AUTOSIZE); imshow("Original - 原始图像", image); namedWindow("Gray - 灰度图像", WINDOW_AUTOSIZE); imshow("Gray - 灰度图像", gray); waitKey(0); destroyAllWindows(); return 0; } - 灰度图像二值化处理
逐个图像像素扫描操作，灰度值大于等于threshold置为255，小于threshold置为0。还增加了一个trackbar用于调节threshold。int Binarization() { Mat binaImg = Mat::zeros(gray.size(), gray.type()); //uchar threshold_val; int threshold_val; threshold_val = 128; namedWindow("Gray - 灰度图像", WINDOW_AUTOSIZE); imshow("Gray - 灰度图像", gray); namedWindow("Binarization", WINDOW_AUTOSIZE); createTrackbar("Threshold", "Binarization", &threshold_val, 255); for (;;) { for (int y = 0; y < gray.rows; y++) { for (int x = 0; x < gray.cols; x++) { //saturate_cast<uchar>(); binaImg.at<uchar>(y, x) = gray.at<uchar>(y, x) >= threshold_val ? 255 : 0; } } imshow("Binarization", binaImg); if (waitKey(10)!=255) { destroyAllWindows(); break; } } return 0; } - 灰度图像的线性变换
逐个图像像素扫描操作，将图像中的灰度级重新映射到minValue到maxValue的范围中。这两个值可以小于0或大于255。被映射到[0,255]区间之外的灰度级会经过saturate_cast<>()函数置为0或255。LinearTransProcessing函数用于进行线性变换，LinearTrans函数用于显示，以及增加一个调节映射区间的trackbar。函数接口仿照OpenCV的API的一般写法写的。int LinearTransProcessing(Mat src, Mat dst, int minValue, int maxValue) { double srcMin, srcMax; minMaxLoc(src, &srcMin, &srcMax, 0, 0); /*参数检查*/ if (dst.type() != CV_8U) { cout << "type error at LinearTrans" << endl; return -1; } if (dst.size() != src.size()) { cout << "size error at LinearTrans: dst.size() != src.size()" << endl; return -2; } if (srcMin == srcMax) { dst = src.clone(); return 0; } if (src.type() == CV_8U) { for (int y = 0; y < src.rows; y++) { for (int x = 0; x < src.cols; x++) { dst.at<uchar>(y, x) = saturate_cast<uchar>(minValue + int((double(src.at<uchar>(y, x)) - srcMin) / (srcMax - srcMin)*double(maxValue - minValue) + 0.5));//int(double + 0.5)用来四舍五入 } } } else if (src.type() == CV_64F) { for (int y = 0; y < src.rows; y++) { for (int x = 0; x < src.cols; x++) { dst.at<uchar>(y, x) = saturate_cast<uchar>(minValue + int((src.at<double>(y, x) - srcMin) / (srcMax - srcMin)*double(maxValue - minValue) + 0.5)); } } } else { cout << "type error at LinearTrans" << endl; return -1; } return 0; } int LinearTrans() { Mat linearImg = Mat::zeros(gray.size(), gray.type()); int max = 255, min = 0; int baseBias = 255; int maxBias = baseBias, minBias = baseBias; namedWindow("Gray - 灰度图像", WINDOW_AUTOSIZE); imshow("Gray - 灰度图像", gray); namedWindow("LinearTrans", WINDOW_AUTOSIZE); createTrackbar("最大值", "LinearTrans", &max, 255); createTrackbar("最小值", "LinearTrans", &min, 255); createTrackbar("maxBias+baseBias", "LinearTrans", &maxBias, 2*baseBias); createTrackbar("minBias+baseBias", "LinearTrans", &minBias, 2*baseBias); for (;;) { LinearTransProcessing(gray, linearImg, min + minBias - baseBias, max + maxBias - baseBias); imshow("LinearTrans", linearImg); if (waitKey(10) != 255) { destroyAllWindows(); break; } } return 0; } - 灰度图像的对数变换
逐个图像像素扫描操作，按对数变换公式计算。如果通过调节系数c使得对数变换后的灰度级能对应到[0,255]区间内，那么底数的变化不会影响变换结果，对数变换将始终遵循：
int LogTrans() { double c = 255 / (log(1 + 255));//以e为底时的系数45.99,对应0-255的映射 Mat logImg = Mat::zeros(gray.size(), gray.type()); for (int y = 0; y < gray.rows; y++) { for (int x = 0; x < gray.cols; x++) { logImg.at<uchar>(y, x) = saturate_cast<uchar>(c*log(1 + gray.at<uchar>(y, x))); } } namedWindow("Gray - 灰度图像", WINDOW_AUTOSIZE); imshow("Gray - 灰度图像", gray); namedWindow("LogTrans", WINDOW_AUTOSIZE); imshow("LogTrans", logImg); waitKey(0); destroyAllWindows(); return 0; } - 灰度图像的伽马变换
逐个图像像素扫描操作，按伽马变换公式$ s = cr^\gamma $计算。指数可以通过trackbar调节，调节范围0.1 - 1 - 10，分别对应较小和较大灰度级的拉伸。int GamaTrans() { Mat gamaImg = Mat::zeros(gray.size(), gray.type()); //uchar threshold_val; double c; double gama, gama_pre; int gama_linear=100; int max_gama_linear = 200; gama = 1; namedWindow("Gray - 灰度图像", WINDOW_AUTOSIZE); imshow("Gray - 灰度图像", gray); namedWindow("GamaTrans", WINDOW_AUTOSIZE); createTrackbar("指数调节", "GamaTrans", &gama_linear, max_gama_linear); for (;;) { gama_pre = gama; gama = (double)pow(10, (double)((double)(gama_linear - max_gama_linear / 2) / (double)(max_gama_linear / 2))); c = (double)255 / (pow(255, (double)gama)); if (fabs(gama - gama_pre) > 0.000001) { cout << "此时指数为：" << gama << endl; } for (int y = 0; y < gray.rows; y++) { for (int x = 0; x < gray.cols; x++) { gamaImg.at<uchar>(y, x) = saturate_cast<uchar>(c*pow(gray.at<uchar>(y, x), (double)gama)); } } imshow("GamaTrans", gamaImg); if (waitKey(10) != 255) { destroyAllWindows(); break; } } return 0; } - 彩色图像的补色变换
这种变换是直接用255减去BGR通道的值，得到的图像色调与原来相反，但饱和度和亮度也随之变化。int ComplementaryColorTrans() { Mat complementaryColorImg = Mat::zeros(image.size(), image.type()); for (int y = 0; y < image.rows; y++) { for (int x = 0; x < image.cols; x++) { complementaryColorImg.at<Vec3b>(y, x)[0] = 255 - image.at<Vec3b>(y, x)[0]; complementaryColorImg.at<Vec3b>(y, x)[1] = 255 - image.at<Vec3b>(y, x)[1]; complementaryColorImg.at<Vec3b>(y, x)[2] = 255 - image.at<Vec3b>(y, x)[2]; } } namedWindow("Original - 原始图像", WINDOW_AUTOSIZE); imshow("Original - 原始图像", image); namedWindow("Complementary Color Trans", WINDOW_AUTOSIZE); imshow("Complementary Color Trans", complementaryColorImg); waitKey(0); destroyAllWindows(); return 0; } - 彩色图像色调取反变换
这种变换是将BGR图像转换成HSV图像，其中H通道的取值范围是[0,179]，对应360度色度盘。要注意色度取反不是用179去减，而是分小于和大于90两种情况。H通道和V通道不变，最后得到的图像饱和度和亮度都得到了保留，与前一个实验结果形成对比。
int HueInvertColorTrans() { vector<Mat> HSVchannels; split(hsv, HSVchannels); Mat hChannel, sChannel, vChannel; hChannel = HSVchannels.at(0); sChannel = HSVchannels.at(1); vChannel = HSVchannels.at(2); /* show the full color situation */ //Mat fullTable = 255 * Mat::ones(1, 256, CV_8U); //LUT(sChannel, fullTable, sChannel); //LUT(vChannel, fullTable, vChannel); //HSVchannels.at(1) = sChannel; //HSVchannels.at(2) = vChannel; //Mat fullColorImage; //merge(HSVchannels, fullColorImage); //cvtColor(fullColorImage, fullColorImage, CV_HSV2BGR); //namedWindow("Full Color Image", WINDOW_AUTOSIZE); //imshow("Full Color Image", fullColorImage); /* LUT函数快速扫描赋值 */ Mat hueInvertColorTable = Mat::zeros(1, 256, CV_8U);//这里必须是256，不然会报中断错误 uchar *p = hueInvertColorTable.ptr(); for (int i = 0; i < 90; i++) { p[i] = i + 90; } for (int i = 90; i < 180; i++) { p[i] = i - 90; } LUT(hChannel, hueInvertColorTable, hChannel); HSVchannels.at(0) = hChannel; Mat hueInvertColorImg; merge(HSVchannels, hueInvertColorImg); cvtColor(hueInvertColorImg, hueInvertColorImg, CV_HSV2BGR); namedWindow("Original - 原始图像", WINDOW_AUTOSIZE); imshow("Original - 原始图像", image); namedWindow("Hue Invert Color Image", WINDOW_AUTOSIZE); imshow("Hue Invert Color Image", hueInvertColorImg); waitKey(0); destroyAllWindows(); return 0; } - 摄像头视频图像二值化
这个很有意思，是开启摄像头后对每帧图像的BGR通道进行二值化，每个通道的阈值有两个，一个low一个high，在区间内的灰度级为255，区间外为0。三个通道一共六个阈值由trackbar进行调节。通过这种方法可以对特定颜色的目标进行识别，测试的时候效果还不错，通过调节trackbar可以锁定某些特定颜色的物体。因为自己逐个像素扫描太慢没法完全满足摄像头图像的实时要求，所以取阈值直接用了inRange函数。int CapThreshold() { int low_r = 100, low_g = 100, low_b = 100; int high_r = 200, high_g = 200, high_b = 200; Mat frame, frame_threshold; VideoCapture cap(0); namedWindow("Video Capture", WINDOW_NORMAL); namedWindow("Binary Image", WINDOW_AUTOSIZE); namedWindow("Binary Image Adjust", WINDOW_NORMAL); for (;;) { cap >> frame; if (frame.empty()) break; createTrackbar("Low R", "Binary Image Adjust", &low_r, 255); createTrackbar("High R", "Binary Image Adjust", &high_r, 255); createTrackbar("Low G", "Binary Image Adjust", &low_g, 255); createTrackbar("High G", "Binary Image Adjust", &high_g, 255); createTrackbar("Low B", "Binary Image Adjust", &low_b, 255); createTrackbar("High B", "Binary Image Adjust", &high_b, 255); inRange(frame, Scalar(low_b, low_g, low_r), Scalar(high_b, high_g, high_r), frame_threshold); imshow("Video Capture", frame); imshow("Binary Image", frame_threshold); if (waitKey(10) != 255) { destroyAllWindows(); break; } } destroyAllWindows(); return 0; }
实验中的问题
包括在实验中遇到的问题，以及解决问题的方法
- 环境配置问题，对于OpenCV 3.2.0,链接器的附加依赖项只需opencv_world320d.lib即可。64位机器要配置x64属性并在x64下Debug
- 数据结构问题，opencv有大量自己的数据结构，需要查看官方文档了解其意义和用法。
- 图像操作问题，实际上也包含了数据结构问题，有许多不同类型相同的功能函数，比如cvLoadImage和imread两个函数都是读取图像，前者返回的是指向图像的指针，后者返回的是图像对象，类似的情况有很多，如cvCvtGray和CvtGray等等，网络上不同的示例有不同的用法，增加了一定困难，百度还是不如直接看官方文档。
- 有些比较早的参考资料使用了Iplmage等指针对图像进行操作。Iplmage是C语言的数据结构，opencv 2以后引入了新的C++接口，多是对于对象的操作，很多针对OpenCV 2以后版本的资料上都建议用新的数据结构和函数