#include <stim/image/image.h>
#include <cmath>
#include <stim/visualization/colormap.h>
#include <stim/image/image_contour_detection.h>
#include <sstream>


void array_multiply(float* lhs, float rhs, unsigned int N);
void array_add(float* ptr1, float* ptr2, float* sum, unsigned int N);

/// This function evaluates the cPb given an multi-channel image

/// @param img is the multi-channel image
/// @param r is an array of radii for different scaled discs(filters)
/// @param alpha is is an array of weights for different scaled discs(filters)
/// @param s is the number of scales

stim::image<float> cPb(stim::image<float> img, int* r, float* alpha, int s){

	unsigned int w = img.width();              // get the width of picture
	unsigned int h = img.height();             // get the height of picture
	unsigned int c = img.channels();		   // get the channels of picture


	stim::image<float> cPb(w, h, 1);               // allocate space for cPb
	unsigned size = cPb.size();                    // get the size of cPb
	memset ( cPb.data(), 0, size * sizeof(float)); // initialize all the pixels of cPb to 0


	unsigned int N = w * h;						// get the number of pixels
	int sigma_n = 3; 							// set the number of standard deviations used to define the sigma

	std::ostringstream ss;                      // (optional) set the stream to designate the test result file

	stim::image<float> temp;                    // set the temporary image to store the addtion result

	for (int i = 0; i < c; i++){
		for (int j = 0; j < s; j++){

			ss << "data_output/cPb_slice"<< i*s + j << ".bmp";  // set the name for test result file (optional)
			std::string sss = ss.str();
		    
			// get the gaussian gradient by convolving each image slice with the mask
			temp = Pb(img.channel(i), r[i*s + j], sigma_n);

			// output the test result of each slice (optional) 
			//stim::cpu2image(temp.data(), sss, w, h, stim::cmBrewer);

			// multiply each gaussian gradient with its weight
			array_multiply(temp.data(), alpha[i*s + j], N);

			// add up all the weighted gaussian gradients
			array_add(cPb.data(), temp.data(), cPb.data(), N);

			ss.str("");   //(optional) clear the space for stream

		}	
	}

	float max = cPb.maxv();						// get the maximum of cPb used for normalization
	array_multiply(cPb.data(), 1/max, N);		    // normalize the cPb

	// output the test result of cPb (optional) 
	//stim::cpu2image(cPb.data(),  "data_output/cPb_0916.bmp", w, h, stim::cmBrewer);

	return cPb;
}