tthis right way to find

    a=sqrt(0.5)*randn(400)
    snr=1/0.5     %0.5 is variance  and signal=1 consider as normalized
    snrdb=10*log(snr)

is this right way to find actual snr wnen i add 'a' marix as noise in a original image.(becuase i needs to add different variance noise in original images and comparing result)

Answer 1

[Matt J]

It's the right way assuming the "original image" satisfies signal=1, however it is that you're measuring that.

[vipul utsav]

i have not measured signal of original image,but i add different noise matrix of different variance in it. so i consider noise=variance of original image and signal=1 (not a measure)

[Matt J]

Well, then it doesn't sound like you're making an SNR calculation of a kind that will have any practical value. SNR stands for Signal-to-Noise Ratio. It's therefore not something you can meaningfully compute based on the noise strength alone. You need to know the strength of your signal, too.

[Image Analyst]

For background, Matt, he has asked essentially this same question 14 times (yes, fourteen) times this month as separate postings - how to get the signal-to-noise ratio when you don't know the signal and you don't know the noise, all you know is the actual noisy signal and sometimes what specific SNR you want. See his profile.

[Matt J]

Yes, I've seen the similarity in his questions. Wasn't sure they were identical, though.