accessing row and col pixel array

edit

You should use mask :

    Mat image=imread("f:/lib/opencv/samples/data/lena.jpg",IMREAD_GRAYSCALE);
    Mat mask = image<100;
    image.setTo(255,mask);
    imshow("r",image);
    waitKey();

if you want to change a gray level in red you have to convert your image first

    Mat imageColor;
    cvtColor(image, imageColor,CV_GRAY2BGR);
    imageColor.setTo(Scalar(0,0,255),mask);
    imshow("rc", imageColor);
    waitKey();

If your image is in color use inRange :

    Mat image = imread("f:/lib/opencv/samples/data/lena.jpg", IMREAD_COLOR);
    Mat mask;
    inRange(image,Vec3b(0,0,0),Vec3b(100,100,100),mask);
    image.setTo(Vec3b(0,0,255), mask);
    imshow("r", image);
    waitKey();

First: your image is in RGB (CV_8UC3) and you treat it as a greyscale image. That's why it goes only to 1/3 of your row. RGB images contain 3xCols bytes (uchar values) per row.

So either you go to 3*cols in the second for loop, or use Vec3b variables.

With Vec3b variables it's also simple to set it in red:

Vec3b red=Vec3b(0,0,25);
...
    //in the second for loop
    Vec3b pixel=image.at<Vec3b>(j, i)
    if ( (pixel[0]+pixel[1]+pixel[2]) /3 < 100) //change color less than 100
    {
        image.at<Vec3b>(j, i) = red; // make it red
    }

Second: if you want to process greyscale images, be sure they are CV_8UC1 type. Convert it if needed.