逍遙文工作室

透視變換相當值得探討，之前稍微研究了透視投影和透視縮短，還在計算機圖學課上寫了透視投影的結果，如今會了OpenCV函式庫，直接呼叫來玩玩更加有趣～

程式碼其實不會很難，只要有提供API，就先不用管裡頭怎麼實做，當然我要的就是效果如何囉：Ｐ

/**
	Theme: Perspective Transform
	compiler: Visual C++ 2010 with OpenCV 2.4.3
	Date: 102/03/22
	Author: HappyMan
	Blog: https://cg2010studio.wordpress.com/
*/
#include <iostream>
#include <opencv2/core/core.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/highgui/highgui.hpp>
const int nOffset=200; // 設定位移幅度

int main(int argc, char *argv[]){
  cv::Mat src_img = cv::imread("happyman.jpg", 1);
  if(src_img.empty()) return -1;
  // 設定變換[之前]與[之後]的坐標 (左上,左下,右下,右上)
  cv::Point2f pts1[] = {cv::Point2f(0,0),cv::Point2f(0,src_img.rows),cv::Point2f(src_img.cols,src_img.rows),cv::Point2f(src_img.cols,0)};
  cv::Point2f pts2[] = {cv::Point2f(0,0),cv::Point2f(0+nOffset,src_img.rows),cv::Point2f(src_img.cols-nOffset,src_img.rows),cv::Point2f(src_img.cols,0)};
  // 透視變換行列計算
  cv::Mat perspective_matrix = cv::getPerspectiveTransform(pts1, pts2);
  cv::Mat dst_img;
  // 變換
  cv::warpPerspective(src_img, dst_img, perspective_matrix, src_img.size(), cv::INTER_LINEAR);
  // 繪製坐標變換[之前]與[之後]的示意圖
  cv::line(src_img, pts1[0], pts1[1], cv::Scalar(255,255,0), 2, CV_AA);
  cv::line(src_img, pts1[1], pts1[2], cv::Scalar(255,255,0), 2, CV_AA);
  cv::line(src_img, pts1[2], pts1[3], cv::Scalar(255,255,0), 2, CV_AA);
  cv::line(src_img, pts1[3], pts1[0], cv::Scalar(255,255,0), 2, CV_AA);
  cv::line(src_img, pts2[0], pts2[1], cv::Scalar(255,0,255), 2, CV_AA);
  cv::line(src_img, pts2[1], pts2[2], cv::Scalar(255,0,255), 2, CV_AA);
  cv::line(src_img, pts2[2], pts2[3], cv::Scalar(255,0,255), 2, CV_AA);
  cv::line(src_img, pts2[3], pts2[0], cv::Scalar(255,0,255), 2, CV_AA);

  cv::namedWindow("src", CV_WINDOW_AUTOSIZE|CV_WINDOW_FREERATIO);
  cv::namedWindow("dst", CV_WINDOW_AUTOSIZE|CV_WINDOW_FREERATIO);
  cv::imshow("src", src_img);
  cv::imshow("dst", dst_img);
  cv::waitKey(0);
}

執行結果顯示兩張影像：

第一張是原始圖加上變換示意圖。

第二張是透視變換後的結果圖。

關鍵函式：warpPerspective～Applies a perspective transformation to an image.

Parameters:

1. src – input image.
2. dst – output image that has the size dsize and the same type as src .
3. M – 3×3 transformation matrix.
4. dsize – size of the output image.
5. flags – combination of interpolation methods (INTER_LINEAR or INTER_NEAREST) and the optional flag WARP_INVERSE_MAP, that sets M as the inverse transformation (src -> dst).
6. borderMode – pixel extrapolation method (BORDER_CONSTANT or BORDER_REPLICATE).
7. borderValue – value used in case of a constant border; by default, it equals 0.

• 沒有回傳值。

對我來說，最重要的會是它所使用的內插法，當然想要結果影像品質較好的話，所花費的計算量就相對較高，也許執行單一影像看不出來，然而當有一連串影像就會非常明顯！特別是針對影片而言～

interpolation –

• INTER_NEAREST – a nearest-neighbor interpolation
• INTER_LINEAR – a bilinear interpolation (used by default)
• INTER_CUBIC – a bicubic interpolation over 4×4 pixel neighborhood
• INTER_AREA – resampling using pixel area relation. It may be a preferred method for image decimation, as it gives moire’-free results. But when the image is zoomed, it is similar to the INTER_NEAREST method.
• INTER_LANCZOS4 – a Lanczos interpolation over 8×8 pixel neighborhood

關鍵函式：getPerspectiveTransform～Calculates a perspective transform from four pairs of the corresponding points.

Parameters:

1. src – Coordinates of quadrangle vertices in the source image.
2. dst – Coordinates of the corresponding quadrangle vertices in the destination image.

• 回傳3×3矩陣。

參考：透視変換を行う、warpPerspective、getPerspectiveTransform。