How can I segment an Image based on gaussian mixture?

I have an image with four different kinds of colors - dark blue, light blue, an even lighter blue - background and white - background. I'd like to extract only the first two categories. This is what I tried.

    bG= mat2gray(image);
    his=imhisthistogram(bG);
    sai = gmdistribution.fit(his,4);
    m1=sai.mu(1);
    v1=sai.Sigma(1);
    m2=sai.mu(2);
    v2=sai.Sigma(2);

    imt=bG<m2+v2; % to get all pixels in original image with less than m2+v2
    figure;
    imagesc(~imt);

I get an Image with all pixels 1. The bG image matrix had values less than 1 but the mean and variance of each gaussian was way higher.

From what I understood, bG is an image matrix and hence will have intensity values. so all the values here are less than 1.

but the means of four guassians are: 0.8604 0.976 2.9182 6.6377. These don't seem to be intensities and so when I tried to get all the pixels in the image less than second mean, all the pixels passed the test. There is something that I am missing here. Can any one help me segment dark blue and light blue cells based on Gaussian mixture. Any help is appreciated!