Competing algorithms for SNR of a image. Which is better?

Is either of these methods preferable to the other for finding SNR of a image?

Method 1:

signalImage1 = double('crop image of L2S11.png');
 noisyImage = imnoise(signalImage1,'salt & pepper',0.02);

   noiseOnlyImage = double(noisyImage) - signalImage1;
  SNR = mean2(signalImage1 ./ noiseOnlyImage );

Method 2:

m=mean2('crop image of  L2S11.png');
d=double('crop image of  L2S11.png');
 sd=std(sd);
 SNR=m/sd;

Method 3:(this is giving answer in double)
I=imread('crop image of  L2S11.png');
 sd=im2double(I);
 m=mean2('crop image of  L2S11.png');
sd1=std(sd);
 SNR=m/sd1;

Answer 1

[Jonas Reber]

I would go for method 2 (3 is the same?) but wouln't one calculate SNR = mean2(im)/std(im(:))?

[Jim]

thank u for ur answer

but method 2 gives the answer as 2.326

and method 3 gives the answer in double

which answer we have to take

[Jonas Reber]

If I do the following:

i = imread('cameraman.tif'); % load sample image

imd = im2double(i);

imdd = double(i);

snr1 = mean2(imd)./std(imd(:));

snr2 = mean2(imdd)./std(imdd(:));

I twice (snr1 and snr2) get 1.9044, don't you?