prove stegano existance with PSNR/MSB

[nunjimmimya]

hi
as the title it implies

i would like to know how to compare original image and stegano image with PSNR or MSB effectively
it is because i tried to compare original image and stegano image with PSNR and the result is awkward

as far a i understand, PSNR value ranging from 0 - 100 and stegano image is consider readable if PSNR value is above 30
but my value is 110

attach is result code from command line

Code: Select all

compare -verbose -metric PSNR Desktop/kami.png Desktop/stegano.png Desktop/kami_dif.png
Desktop/kami.png PNG 640x480 640x480+0+0 8-bit sRGB 348KB 0.030u 0:00.019
Desktop/stegano.png PNG 640x480 640x480+0+0 16-bit sRGB 454KB 0.040u 0:00.049
Image: Desktop/kami.png
  Channel distortion: PSNR
    red: 110.865
    green: 110.922
    blue: 110.916
    all: 110.901

appreciate your help

thank you

[fmw42]

PSNR is expressed as decibels and so can in principle be larger than 100. But even 100 is a huge difference

see
http://www.imagemagick.org/Usage/compare/#statistics

You might be better using rmse (root mean squared error).


If I do

compare -metric psnr rose: rose_message.png null:
104.287

my guess is that is really -104.287 (amplitude not power) decibels as there should be very little difference

compare -metric rmse rose: rose_message.png null:
0.400052 (6.1044e-06)

The 6.1e-06 is a difference in the range of 0 to 1. So this is 0.0006 % error

see http://en.wikipedia.org/wiki/Decibel


I suspect the only issue is that it should be negative, which I would suspect is an oversight.

[nunjimmimya]

requirement for my project is whether use PSNR or MSB(Most Significant Bit)
if i not wrong, the script will generate MSB (correct me if I'm wrong) when 2 image are being compared

i think i will go for MSB

thank you for your time

PSNR is expressed as decibels and so can in principle be larger than 100. But even 100 is a huge difference

see
http://www.imagemagick.org/Usage/compare/#statistics

You might be better using rmse (root mean squared error).


If I do

compare -metric psnr rose: rose_message.png null:
104.287

my guess is that is really -104.287 (amplitude not power) decibels as there should be very little difference

compare -metric rmse rose: rose_message.png null:
0.400052 (6.1044e-06)

The 6.1e-06 is a difference in the range of 0 to 1. So this is 0.0006 % error

see http://en.wikipedia.org/wiki/Decibel


I suspect the only issue is that it should be negative, which I would suspect is an oversight.

[fmw42]

I do not believe that there is a MSB error measure, only


MAE ..... Mean absolute error (average channel error distance)
MSE ..... Mean squared error (averaged squared error distance)
RMSE .... (sq)root mean squared error -- IE: sqrt(MSE)

or

PAE ..... Peak Absolute Error (within a channel, for 3D color space)
PSNR .... Peak Signal to noise ratio (used in image compression papers)

see error measures at http://www.imagemagick.org/Usage/compare/#statistics