I am trying to create a Mat object with a float type data. Mat kernelxx = Mat( HEIGHT, WIDTH, CV_32FC1 );
But what I receive is unsigned char data of type *uchar::Mat::data. Can any one please help me. I also tried using create constructor as well as convertTo() method to make a float data. But all of them create only *uchar data.
no worries, your Mat creation is correct.
float *fp = kernelxx.ptr<float>(0); //will give you the pointer to the 1st row.
(the underlying data buffer is still uchar*)
Thanks for the answer. But it still didn't solve my purpose. My underlying Mat object still has uchar data. I need a mat object with float data.
Mat kernelxx = Mat::ones( HEIGHT, WIDTH, CV_32FC1 )/HEIGHT*WIDTH; //this creates data with zeros This is weird, as Mat consturctor supports creation of float data.
Mat src=imread("variance8inv.png"); Mat dst(imageHeight, imageWidth, CV_32FC1); filter2D(src, dst, -1 , kernelxx, Point( -1, -1 ), 0, BORDER_DEFAULT );
"this creates data with zeros" - i don't think so.
Mat kernelxx = Mat::ones( 5,6, CV_32FC1 ); cerr << kernelxx << endl;
[1, 1, 1, 1, 1, 1; 1, 1, 1, 1, 1, 1; 1, 1, 1, 1, 1, 1; 1, 1, 1, 1, 1, 1; 1, 1, 1, 1, 1, 1]
Mat src=imread("variance8inv.png"); // this will load a 3 channel image, you want:
Mat src=imread("variance8inv.png", 0); // load as single gray channel
also: filter2D(src, dst, CV_32FC1 , kernelxx, Point( -1, -1 ), 0, BORDER_DEFAULT );
it will overwrite/realloc dst anyway. maybe you even want to src.convertTo(src, CV_32F); before.
Asked: 2014-05-30 05:03:50 -0600
Seen: 275 times
Last updated: May 30 '14
