using Mat, and recognizing imread function

Question

I have been trying to make a median filter using vivado hls. A person advised me to use the OpenCV imread function, but when I do use it the program doesn't recognize it even though I have included the library of OpenCV. If I don't use imread function and use Mat and cvLoadImage I get an error saying the following:

/../HLSAXIStreamMedian_tb.cpp:109:15: error: missing template arguments before '.' token

If anyone knows how to solve this problem or give me an alternative I would appreciate it. This is the code from my test bench:

#include "HLSAXIStreamMedian.h"

int main(void) {

    spix imageIn[MAX_HEIGHT][MAX_WIDTH]; // May have to malloc these if they're large
    spix imageOut[MAX_HEIGHT][MAX_WIDTH];
    spix imageGold[MAX_HEIGHT][MAX_WIDTH];

    static uint8_t frameIn[MAX_HEIGHT * MAX_WIDTH];

    FILE * imgFile = fopen("C:/img.bin","rb"); // "rb" is important - forces the image to be opened in binary mode.

    fread(frameIn,1,MAX_HEIGHT*MAX_WIDTH,imgFile);

    fclose(imgFile);

    // Organise that data into the required image type.

    spix dataIn[MAX_HEIGHT][MAX_WIDTH];

    for (int y = 0; y < MAX_HEIGHT; y++) {
        for (int x = 0; x < MAX_WIDTH; x++) {
            spix tmp;
            tmp.data = frameIn[y*MAX_WIDTH + x];
            tmp.last = (x == (MAX_WIDTH-1));
            tmp.user = (x == 0) && (y == 0);
            dataIn[y][x] = tmp;
        }
    }

    // spix array is now initialised.
//    Mat gold(1212, 1216, CV_8UC3, Scalar(0,0, 100)); //create an image ( 3 channels, 8 bit image depth, 1212 high, 1216 wide, (0, 0, 100) assigned for Blue, Green and Red plane respectively. )
//    imageGold = gold;
    IplImage* test = cvLoadImage("C:/MedianTrial/test_image.PNG");
    test = Mat.dataIn;
    IplImage* gold = cvLoadImage("C:/MedianTrial/gold_output.PNG");
    gold = Mat.imageGold;
    //Mat test;
    //test = imread("C:/MedianTrial/test_image.PNG",in_pix); // Read a sample image. "imread" does not actually exist; read or generate an image using a method of your choice.
    //Mat gold;
    //gold = imread("C:/MedianTrial/gold_output.PNG",imageGold); // Read a known-good image (eg. generated using Matlab's median filter).

    top_median(imageIn,imageOut,1080,1920); // Call the test function.

    for (int i = 0; i < 1080; i++) {
        for (int j = 0; j < 1920; j++) {
            if (imageOut != imageGold) {
                printf("Data mismatch");// at position (%d,%d): %d != %d\n",i,j,imageOut[i][j],imageGold[i][j]);
                return -1;
            }
        }
    }
}

This is the header file:

#define KMED 3 // KMED can be 3, 5, 7
#define KKMED 1 // KKMED == 1 for 3x3 window
#define MIN(x,y) ( (x)>(y) ? (y) : (x) )
#define MAX(x,y) ( (x)>(y) ? (x) : (y) )
#include "opencv/cxcore.h"
#include <opencv2/core/core.hpp>
#include "opencv2/highgui/highgui.hpp"
//opencv_core
//opencv_imgcodecs
//opencv_highgui
#include <iostream>
#include "hls_video.h"
#include "hls_opencv.h"
#include <ap_int.h>
#include "ap_axi_sdata.h"
#include <stdio.h>
#include <string.h>
#include <hls_stream.h>
#include <stdlib.h>
using namespace hls;
//using namespace cv;
//using namespace std;
#include <ap_axi_sdata.h>
typedef ap_axis<8,2,5,6> spix;
#ifndef GRAY11
typedef unsigned char pix_t; // 8-bit per pixel
#else
#include <ap_int.h>
typedef ap_int<11> pix_t; // 11-bit per pixel
#endif
#define MAX_HEIGHT 1080
#define MAX_WIDTH 1920
void top_median(spix in_pix[MAX_HEIGHT][MAX_WIDTH],
spix out_pix[MAX_HEIGHT][MAX_WIDTH],
short int height, short int width);
//spix in_pix[MAX_HEIGHT][MAX_WIDTH];
//spix out_pix[MAX_HEIGHT][MAX_WIDTH];
//spix out_pixtb[MAX_HEIGHT][MAX_WIDTH];
//short int height;
//short int width;
pix_t Median(pix_t window[KMED*KMED]);
    enter code here
    enter code here

Answer 1

I can't give you complete solution as I don't have hls library but can point you to right direction.

1. Don't use using namespace ....

2. Load image using cvLoadImage() or cv::imread().

3. a. (if cvLoadImage) Convert IplImage* to hls::Mat using IplImage2hlsMat().
   b. (if imread) Convert cv::Mat to hls::Mat using cvMat2hlsMat().
   It is important to convert to hls::Mat not to cv::Mat.

4. Create hls::Window and initialize it with your filter for ex.
   [ 1, 1, 1 ]
   [ 1, 1, 1 ]
   [ 1, 1, 1 ]

5. Use hls::Filter2D with your hls::Mat and hls::Window to apply median filter.

Refs:

Article how to create filter in OpenCV
HLS doc