在cv :: Mat中将float转换为uchar的偏移量

Question

In process of speeding up some processes (can't name them, sorry), I tried to create a

cv::Mat_<uchar> discretization;

Now when I get a depth map in float

cv::Mat_<float> depth_map;
discretization = depth_map / resolution_mtr;

where resolution_mtr is a float. Its value is 0.1 currently. When I do this, for a value say, 0.48 in depth map , I get the discretization value of 5. My understanding says it should be 4 . I guess it is round off to nearest uchar. Is there a way out of this without getting into for loop ? Basically I want to use floor values in discretization and not round off .

Answer 1

为什么不定义继承的类CvNoRoundMat并覆盖其operator +？

Answer 2

You can just subtract 0.5 from the result.

This code

float resolution_mtr = 0.1;
float vals[] = {0.48, 0.4, 0.38, 0.31};
cv::Mat_<float> depth_map(1,4,vals);
cv::Mat_<uchar> discretization( depth_map  / resolution_mtr - 0.5);

std::cout << "depth_map: " << depth_map << std::endl;
std::cout << "discretization: " << discretization << std::endl;  

will give you following results:

depth_map: [0.47999999, 0.40000001, 0.38, 0.31]
discretization: [4, 4, 3, 3]