source: trunk/src/transform_field/ima_noise_rms.m @ 803

% 'ima_noise_rms': gives the variance of relative noise by difference to the
% filered image in ppm (part per million) (for grey scale image)

%------------------------------------------------------------------------
%%%%  Use the general syntax for transform fields with a single input and parameters %%%%
% OUTPUT: 
% DataOut:   output field structure 

%INPUT:
% DataIn:  input field structure
% Param: matlab structure whose field Param.TransformInput contains the filter parameters
%-----------------------------------

%-------------------------------------
function DataOut=ima_noise_rms(DataIn,Param)

%% request input parameters
if isfield(DataIn,'Action') && isfield(DataIn.Action,'RUN') && isequal(DataIn.Action.RUN,0)
    prompt = {'npx';'npy'};
    dlg_title = 'get the filter size in x and y';
    num_lines= 2;
    def     = { '20';'20'};
    if isfield(Param,'TransformInput')&&isfield(Param.TransformInput,'FilterBoxSize_x')&&...
            isfield(Param.TransformInput,'FilterBoxSize_y')
        def={num2str(Param.TransformInput.FilterBoxSize_x);num2str(Param.TransformInput.FilterBoxSize_y)};
    end
    answer = inputdlg(prompt,dlg_title,num_lines,def);
    DataOut.TransformInput.FilterBoxSize_x=str2num(answer{1}); %size of the filtering window
    DataOut.TransformInput.FilterBoxSize_y=str2num(answer{2}); %size of the filtering window
    return
end

DataOut=DataIn; %default

%definition of the cos shape matrix filter
ix=[1/2-Param.TransformInput.FilterBoxSize_x/2:-1/2+Param.TransformInput.FilterBoxSize_x/2];%
iy=[1/2-Param.TransformInput.FilterBoxSize_y/2:-1/2+Param.TransformInput.FilterBoxSize_y/2];%
%del=np/3;
%fct=exp(-(ix/del).^2);
fct2_x=cos(ix/((Param.TransformInput.FilterBoxSize_x-1)/2)*pi/2);
fct2_y=cos(iy/((Param.TransformInput.FilterBoxSize_y-1)/2)*pi/2);
%Mfiltre=(ones(5,5)/5^2);
Mfiltre=fct2_y'*fct2_x;
Mfiltre=Mfiltre/(sum(sum(Mfiltre)));%normalize filter

Atype=class(DataIn.A);% detect integer 8 or 16 bits
B=filter2(Mfiltre,DataIn.A);
B(B==0)=1; %set to 1 the zero values
C=(double(DataIn.A)-B)./B;
C=filter2(Mfiltre,C.*C);% take variance  integrated in the filtering area 
C=1000000*sqrt(C); % take the root and *10^6 to get an image with integer values (parts per million)
DataOut.A=feval(Atype,C);%transform to the initial image format