Assuming the noise mean has zero bias, just take the mean of the image over the white part of the target and the black part of the target, and take the standard deviation, then divide
meanWhiteGL = mean2(whitePartOfImage);
stdWhiteGL = std(double(whitePartOfImage(:)));
SNRWhite = meanWhiteGL / stdWhiteGL;
Same for the black part of the image. It's possible that the two regions could have different SNRs.
