PSNR计算

purpose_857

im1=imread('lena2.bmp');
3 Q& N1 h; T/ e' u9 l5 n6 g' ~%im1=im2double(im1);
im2=imread('lena_rbhv.bmp');
( l( l, P- n$ \! ?. Q%im2=im2double(im2);
0 k$ |) t% ~. O% ~+ z! C   
if (size(im1))~=(size(im2))
    error('错误：两个输入图象的大小不一致');
end
: O2 z5 e7 {" L4 |) l* x7 \" j, {; G) T* h
# V7 ^: S* F5 R7 e    [m,n] = size(im1);
( k- t0 X% q  P& @    A = double(im1);
8 K7 `  k! z4 \" k$ F    B = double(im2);
    D = sum( sum( (A-B).^2 ) );%||A-B||^2
    MSE = D / (m * n);
5 r8 S4 E) _  R: i8 W+ R% [if  D == 0
$ g" X6 }8 c5 y. c9 c- J% Y$ q    error('两幅图像完全一样');
    PSNR = 200;
else
+ U  p' V" F/ _' C6 e    PSNR = 10*log10( (255^2) / MSE );                                                        
end
& Q- G8 x$ h, A7 S

从网上COPY了下 稍微改了下  我不知道这一行为什么出错  请指教
??? Subscript indices must either be real positive integers or logicals.

+ S% E/ M( w9 S. |7 |Error in ==> PSNR_ERRO at 14
6 n: X$ r% {1 M' g    D = sum( sum( (A-B).^2 ) );%||A-B||^2

I_believe

D = sum( sum( (A-B).^2 ) );指的是什么，为什么有两从sum？

peerless2021

function main()
- R8 |% ?( k/ `) _/ o% w8 T6 Dclc; close all;
I=imread('cameraman.tif'); %用的是原图
I1=imnoise(I,'gaussian',0,0.01);
figure;
4 b4 \9 ~( ~. K/ I( b1 u! Esubplot(1, 2, 1); imshow(I); title('原图像');
2 Y3 e4 l2 K, t; M9 ysubplot(1, 2, 2); imshow(I1); title('加噪声图像');
[PSNR, MSE] = psnr(I, I1)
% 图像峰值信噪比PSNR、均方根误差MSE
" e6 a% F$ |  a  y+ d) Z0 W" _function [PSNR, MSE] = psnr(X, Y)
! e$ D* W: S6 A$ Z& b% 计算峰值信噪比PSNR、均方根误差MSE
" p* v) p6 _+ I" g% 如果输入Y为空，则视为X与其本身来计算PSNR、MSE
$ L0 ^* _$ d# o' k. n( }
if nargin<2
- ~( a2 t5 x% M1 ]    D = X;
else
    if any(size(X)~=size(Y))
0 F$ f9 \+ ^6 }        error('The input size is not equal to each other!');
    end
- y( [, e7 c! x- s7 d    D = X-Y;
end
+ s% L: X+ {# e0 h8 Q  a7 jMSE = sum(D(: ).*D(: ))/prod(size(X));
  B8 a0 X$ ^9 s2 V1 x" K8 N5 yPSNR = 10*log10(255^2/MSE);
function main()
clc; close all;
( y9 D2 L3 T( CI=imread('cameraman.tif'); %用的是原图
I1=imnoise(I,'gaussian',0,0.01);
9 S: W4 n; G( \3 q5 b* e" E2 ifigure;
subplot(1, 2, 1); imshow(I); title('原图像');
, p2 s+ R  ^( c* s* Asubplot(1, 2, 2); imshow(I1); title('加噪声图像');
  ?$ k6 x( @) H8 k9 U8 U[PSNR, MSE] = psnr(I, I1)
% 图像峰值信噪比PSNR、均方根误差MSE
function [PSNR, MSE] = psnr(X, Y)
7 A1 i' F0 \" j9 v* K% 计算峰值信噪比PSNR、均方根误差MSE
. S8 j* s" H; l' o( q% 如果输入Y为空，则视为X与其本身来计算PSNR、MSE
5 m, C1 L$ o- L4 o8 V2 ?( F8 h# y  d
2 X! j; Z, A" ]! e6 l* Q  Kif nargin<2
. ^' Z+ `* p( J% B- _5 S    D = X;
$ q5 Y( F" ^$ v( Yelse
1 [* {: k) @7 M: Z+ K0 k2 E" w    if any(size(X)~=size(Y))
        error('The input size is not equal to each other!');
; Y8 W+ ]& z( q! h- s3 _1 e    end
    D = X-Y;
end
9 q# @* o! u) P. ]0 |MSE = sum(D(: ).*D(: ))/prod(size(X));
6 E& t5 D  a! P& X/ wPSNR = 10*log10(255^2/MSE);
用这个算法算下吧，I为原图 I1为加噪图像
/ u1 H# U7 x- H$ h+ u7 o