Extract hand bones from X-ray image - opencv

Question

I have x-ray image of a hand. I need to extract bones automatically. I can easily segmentate a hand using different techniques. But I need to get bones and using those techniques don't help. Some of the bones are brighter then orthers, so if I use thresholding some of them disapear while others become clearer rising threshold. And I think maybe I should threshold a region of the hand only? Is it possible to threshold ROI that is not a square? O maybe you have any other solutions, advices? Maybe there are some libraries like OpenCV or something for that? Any help would be very great!
Extended:
Raw Image Expected Output

Answer 1

One approach could be to segment the hand and fingers from the image:
And then creating another image with just the hand silhouette:
Once you have the silhouette you can erode the image to make it a little smaller. This is used to subtract the hand from the hand & fingers image, resulting in the fingers:
The code below shows to execute this approach:
void detect_hand_and_fingers(cv::Mat& src);
void detect_hand_silhoutte(cv::Mat& src);
int main(int argc, char* argv[])
{
cv::Mat img = cv::imread(argv[1]);
if (img.empty())
{
std::cout << "!!! imread() failed to open target image" << std::endl;
return -1;
}
// Convert RGB Mat to GRAY
cv::Mat gray;
cv::cvtColor(img, gray, CV_BGR2GRAY);
cv::Mat gray_silhouette = gray.clone();
/* Isolate Hand + Fingers */
detect_hand_and_fingers(gray);
cv::imshow("Hand+Fingers", gray);
cv::imwrite("hand_fingers.png", gray);
/* Isolate Hand Sillhoute and subtract it from the other image (Hand+Fingers) */
detect_hand_silhoutte(gray_silhouette);
cv::imshow("Hand", gray_silhouette);
cv::imwrite("hand_silhoutte.png", gray_silhouette);
/* Subtract Hand Silhoutte from Hand+Fingers so we get only Fingers */
cv::Mat fingers = gray - gray_silhouette;
cv::imshow("Fingers", fingers);
cv::imwrite("fingers_only.png", fingers);
cv::waitKey(0);
return 0;
}
void detect_hand_and_fingers(cv::Mat& src)
{
cv::Mat kernel = cv::getStructuringElement(cv::MORPH_ELLIPSE, cv::Size(3,3), cv::Point(1,1));
cv::morphologyEx(src, src, cv::MORPH_ELLIPSE, kernel);
int adaptiveMethod = CV_ADAPTIVE_THRESH_GAUSSIAN_C; // CV_ADAPTIVE_THRESH_MEAN_C, CV_ADAPTIVE_THRESH_GAUSSIAN_C
cv::adaptiveThreshold(src, src, 255,
adaptiveMethod, CV_THRESH_BINARY,
9, -5);
int dilate_sz = 1;
cv::Mat element = cv::getStructuringElement(cv::MORPH_ELLIPSE,
cv::Size(2*dilate_sz, 2*dilate_sz),
cv::Point(dilate_sz, dilate_sz) );
cv::dilate(src, src, element);
}
void detect_hand_silhoutte(cv::Mat& src)
{
cv::Mat kernel = cv::getStructuringElement(cv::MORPH_ELLIPSE, cv::Size(7, 7), cv::Point(3, 3));
cv::morphologyEx(src, src, cv::MORPH_ELLIPSE, kernel);
int adaptiveMethod = CV_ADAPTIVE_THRESH_MEAN_C; // CV_ADAPTIVE_THRESH_MEAN_C, CV_ADAPTIVE_THRESH_GAUSSIAN_C
cv::adaptiveThreshold(src, src, 255,
adaptiveMethod, CV_THRESH_BINARY,
251, 5); // 251, 5
int erode_sz = 5;
cv::Mat element = cv::getStructuringElement(cv::MORPH_ELLIPSE,
cv::Size(2*erode_sz + 1, 2*erode_sz+1),
cv::Point(erode_sz, erode_sz) );
cv::erode(src, src, element);
int dilate_sz = 1;
element = cv::getStructuringElement(cv::MORPH_ELLIPSE,
cv::Size(2*dilate_sz + 1, 2*dilate_sz+1),
cv::Point(dilate_sz, dilate_sz) );
cv::dilate(src, src, element);
cv::bitwise_not(src, src);
}