OpenCV reading image pixel value - c++

Question

I'm trying very simple thing with opencv but getting error.
I'm just trying read 16bit png image and access to the specific pixel value. I tried many ways but couldn't manage the get value. I'm using OpenCV3.0 on windows8 64bit.
NOTE: while reading image using CV_LOAD_IMAGE_GRAYSCALE is fine, but CV_LOAD_IMAGE_ANYDEPTH rising error. But when i use CV_LOAD_IMAGE_GRAYSCALE my highest pixel is 9, which should be around 2000
I uploaded example image.example image
my example code:
#include "opencv2/core/core.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/highgui/highgui.hpp"
int _tmain(int argc, _TCHAR* argv[])
{
cv::Mat frame = cv::imread("filepath", CV_LOAD_IMAGE_ANYDEPTH );//using CV_LOAD_IMAGE_GRAYSCALE is fine, but CV_LOAD_IMAGE_ANYDEPTH rising error
frame.convertTo(frame, CV_16U);// to be sure... i omitted this part also and same error
double min, max;
cv::Point mloc, mxloc;
cv::minMaxLoc(frame, &min, &max, &mloc, &mxloc);
//i can access min and max values but not the specific pixel value
float zmx = frame.at<unsigned char>(118, 38);//rise error
float zm = frame.at<float>(30,40);//rise error
return 0;
}
Error message:
Unhandled exception at 0x00007FF8EB288A5C in OpenCVTest.exe: Microsoft C++ exception: cv::Exception at memory location 0x000000A47F40F230.
But i think this is misleading error, i check the my image is 320*240, so i'm sure that there is pixel at that location.
I tried with Scalar also, but i'm getting same error

Answer 1

Your biggest problem is that you're trying to access a 16 bpp image, i.e. a Mat of type CV_16U with the wrong data type. You should use frame.at<ushort>(...) in case of single channel 16 bpp image (I suppose this is the case here), or with frame.at<Vec3w>(...) for 3 channels images.
Also you should make sure that you're loading the image properly. Using imread with parameter IMREAD_GRAYSCALE you're converting your image to 8bpp, which is not what you want. You should use IMREAD_ANYDEPTH or IMREAD_UNCHANGED.
Take a look at this code:
#include<opencv2/opencv.hpp>
int main()
{
// Read the image as original bpp
cv::Mat frame = cv::imread("path_to_image", cv::IMREAD_ANYDEPTH);
// Be sure that the image is loaded
if (frame.empty())
{
// No image loaded
return -1;
}
// Be sure that the image is 16bpp and single channel
if (frame.type() != CV_16U || frame.channels() != 1)
{
// Wrong image depth or channels
return -1;
}
double min_val, max_val;
cv::Point min_loc, max_loc;
cv::minMaxLoc(frame, &min_val, &max_val, &min_loc, &max_loc);
// Access values with correct data type
ushort zmx = frame.at<ushort>(max_loc);
return 0;
}

Answer 2

A couple of things you should notice.
First you are defining frame twice.
second, this line float zmx = frame.at<unsigned char>(118, 38); has a couple of issues. you are assigning unsigned char to a float. also you should notice the order is reversed to access the x,y pixel value you call frame.at<unsigned char>(y, x) then best way to assign to a Scalar instead like this
Scalar fmx = frame.at<uchar>(118, 38);
or use Point to avoid confusion
Scalar fmx = frame.at<uchar>(Point(38,118));
last thing, make sure you loaded the image properly and that frame has the image data
UPDATE
I just tested your code and it worked fine (check below), I can't think of anything but not finding the image in the path provided
#include "opencv2/core/core.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/highgui/highgui.hpp"
int main(int argc, char** argv)
{
cv::Mat frame = cv::imread("0FD0X.png", CV_LOAD_IMAGE_GRAYSCALE);
frame.convertTo(frame, CV_16U);// to be sure... i omitted this part also and same error
double min, max;
cv::Point mloc, mxloc;
cv::minMaxLoc(frame, &min, &max, &mloc, &mxloc);
//i can access min and max values but not the specific pixel value
float zmx = frame.at<unsigned char>(118, 38);// no error
float zm = frame.at<float>(30, 40);// no error
std::cout << zmx << std::endl; // out 0
std::cout << min << std::endl; // out 0
std::cout << max << std::endl; // out 9
std::cout << mloc << std::endl; // out [0,0]
std::cout << mxloc << std::endl; // out [125,30]
return 0;
}
Update #2 to access multichannel image you need to access using the Vec3b data type. Also notice the order of the point coordinates. check the following code
#include "opencv2/core/core.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/highgui/highgui.hpp"
int main(int argc, char** argv)
{
cv::Mat frame = cv::imread("0FD0X.png", CV_LOAD_IMAGE_ANYDEPTH);
frame.convertTo(frame, CV_16U);// to be sure... i omitted this part also and same error
double min, max;
cv::Point mloc, mxloc;
cv::minMaxLoc(frame, &min, &max, &mloc, &mxloc);
//i can access min and max values but not the specific pixel value
ushort pValShort = frame.at<ushort>(38, 118);// no error
Vec3b pValVec = frame.at<Vec3b>(38, 118);// no error
Vec3b pValVecPoint = frame.at<Vec3b>(Point(118,38));// no error
std::cout << pValShort << std::endl; // out 2423
std::cout << pValVec << std::endl; // out [166,8,165]
std::cout << pValVecPoint << std::endl; // out [166,8,165]
std::cout << min << std::endl; // out 0
std::cout << max << std::endl; // out 2423
std::cout << mloc << std::endl; // out [0,0]
std::cout << mxloc << std::endl; // out [118,38]
return 0;
}