Merge branch 'master' of gitlab.gwdg.de:irp/holotomotoolbox

examples/template_tomo_analysis_Julia.m

@@ -4,7 +4,6 @@ set(0,'DefaultFigureColor','w');
 
 
 %% parameters of the scan
-spec = 'julia';
 year = '2019';
 runname = '2019_05_02';
 detector = 'argos';
@@ -29,16 +28,15 @@ if ispc
     toolboxPath = 'S:/Projects_X-ray_Imaging/holotomotoolbox/functions';
     astraPath = 'S:/Projects_X-ray_Imaging/ASTRAToolbox/windows';
     prePath = 'J:/';
-    saveDir = fullfile('T:/Julialabor Tomo', runname, filePrefix, 'Slices');
 elseif isunix
-    toolboxPath = '/IRP/AG_Salditt/Projects_X-ray_Imaging/holotomotoolbox/functions';
-    astraPath = '/IRP/AG_Salditt/Projects_X-ray_Imaging/ASTRAToolbox/linux';
-    prePath = '/IRP/Labdata/AG_Salditt/julialab/';
-    saveDir = fullfile('/tmp/', filePrefix);
+    toolboxPath = '/home/AG_Salditt/Projects_X-ray_Imaging/holotomotoolbox/functions';
+    astraPath = '/home/AG_Salditt/Projects_X-ray_Imaging/ASTRAToolbox/linux';
+    prePath = '/home/Labdata/AG_Salditt/julialab/';
 end
 
 % change if not using julia
 dataPath = fullfile(prePath, year, 'julia' , runname, 'detectors', detector, filePrefix);
+saveDir = fullfile(prePath, year, 'julia', runname, 'analysis', filePrefix, 'Slices');
 
 if ~exist(toolboxPath,'dir')
     error(['Assigned path to the holotomotoolbox ''', toolboxPath, ''' does not exist.']);
@@ -155,7 +153,7 @@ parfor indAngle = 1:numAngles
     number = rawnumbers(indAngle);
     imgTmp = zeros(size(raw0,1),size(raw0,2),numAverage);
     for indImage = 1:numAverage
-        imgTmp(:,:,indImage) = imageReader(getFileName(number+(indImage-1)));
+        imgTmp(:,:,indImage) = imageReader(getFileName(number+(indImage-1))); %#ok<PFBNS>
     end
     % average over all images
     raw = median(imgTmp,3);
@@ -246,12 +244,7 @@ z12 = (z02-z01)/1000;
 z_eff = z12/M;
 
 % main energy peak of the x-ray spectrum
-switch lower(spec)
-    case 'mona'
-        E = 17.5;					% molybdenum
-    case 'julia'
-        E = 9.25;					% gallium
-end
+E = 9.25;					% gallium
 
 % wavelength of the main energy peak
 lambda = 12.398/E*1e-10;

functions/auxiliary/fftfreq.m

-function varargout = fftfreq(N, dx, computeMeshgrid)
+function varargout = fftfreq(N, dx)
 % FFTFREQ creates a grid of Fourier-frequencies corresponding to the sampling points
 % of an n-dimensional FFT of given size.
 %
-%     ``varargout = fftfreq(N, dx, computeMeshgrid)``
+%     ``varargout = fftfreq(N, dx)``
 % 
 %
 % Parameters
@@ -12,16 +12,12 @@ function varargout = fftfreq(N, dx, computeMeshgrid)
 % dx : float or tuple of floats, optional
 %     Underyling spacing in real-space, possibly different along each grid-dimension. 
 %     Default = ones(size(N)).
-% computeMeshgrid : tuple of non-negative integers, optional
-%     Return Fourier-frequency grid as meshgrid (case: true) or merely as coordinate vectors
-%     for the different dimensions? Default = false.
 %
 %
 % Returns 
 % ------- 
 % varargout : tuple of numerical arrays
 %     numel(N) vectors of length N(1), N(2),... that span the Fourier-grid 
-%     OR, if computeMeshgrid == true, the meshgrid corresponding to these vectors
 %
 % See also
 % --------
@@ -64,19 +60,18 @@ dx = dx(:).' .* ones([1,numel(N)]);
 ndim = numel(N);
 
 % Grid in Fourier space
+xi = cell(1, ndim);
 for jj = 1:ndim
     xi{jj} = ifftshift( ( -floor(0.5*N(jj)) : floor(0.5*(N(jj)-1)) ).' * ( 2*pi / (N(jj)*dx(jj)) ) );
 end
 
-% Optionally assemble ndgrid (occupies more memory!)
-if nargin == 3 && computeMeshgrid
-    [xi{:}] = ndgrid(xi{:});
-elseif ndim > 1
+if ndim > 1
     for jj = 1:ndim
         xi{jj} = reshape( xi{jj}, [ones([1,jj-1]), N(jj), ones([1,ndim-jj])] );
     end
 end
 
+
 % Unpack cell-array
 varargout = xi;
 

functions/auxiliary/fftfreqNormSq.m

@@ -47,8 +47,8 @@ if nargin < 2
     dx = ones(size(N));
 end
 
-xi = cell([numel(N),1]);
-[xi{:}] = fftfreq(N, dx, false);
+xi = cell(numel(N),1);
+[xi{:}] = fftfreq(N, dx);
 XiSq = xi{1}.^2;
 for dim = 2:numel(xi)
     XiSq = XiSq + xi{dim}.^2;

functions/fresnelPropagation/fresnelPropagationKernel.m

@@ -138,7 +138,7 @@ function kernel = fresnelPropagationKernel_fourier(N, fresnelNumbers, gpuFlag)
 
 ndim = numel(N);
 xi = cell([ndim,1]);
-[xi{:}] = fftfreq(N, 1, false);
+[xi{:}] = fftfreq(N, 1);
 kernel = 1;
 for dim = 1:ndim
     xi{dim} = gpuArrayIf(xi{dim}, gpuFlag);
@@ -154,7 +154,7 @@ function kernel = fresnelPropagationKernel_chirp(N, fresnelNumbers, gpuFlag)
 
 ndim = numel(N);
 x = cell([ndim,1]);
-[x{:}] = fftfreq(N, 2*pi./N, false);    % fft-shifted grid with spacing 1
+[x{:}] = fftfreq(N, 2*pi./N);    % fft-shifted grid with spacing 1
 kernel = prod(sqrt(abs(fresnelNumbers)) .* exp((-1i*pi/4) .* sign(fresnelNumbers)));
 for dim = 1:ndim
     x{dim} = gpuArrayIf(x{dim}, gpuFlag);
@@ -170,7 +170,7 @@ function kernel = fresnelPropagationKernel_chirpLimited(N, fresnelNumbers, gpuFl
 
 ndim = numel(N);
 x = cell([ndim,1]);
-[x{:}] = fftfreq(N, 2*pi./N, false);    % fft-shifted grid with spacing 1
+[x{:}] = fftfreq(N, 2*pi./N);    % fft-shifted grid with spacing 1
 kernel = prod(sqrt(abs(fresnelNumbers)) .* exp((-1i*pi/4) .* sign(fresnelNumbers)));
 for dim = 1:ndim
     x{dim} = gpuArrayIf(x{dim}, gpuFlag);
@@ -196,7 +196,7 @@ function kernel = fresnelPropagationKernel_pixel2pixel(N, fresnelNumbers, gpuFla
 
 ndim = numel(N);
 x = cell([ndim,1]);
-[x{:}] = fftfreq(N, 2*pi./N, false);    % fft-shifted grid with spacing 1
+[x{:}] = fftfreq(N, 2*pi./N);    % fft-shifted grid with spacing 1
 kernel = 1;
 for dim = 1:ndim
     x{dim} = gpuArrayIf(x{dim}, gpuFlag);

functions/imageProcessing/analysis/angularAverage.m

@@ -28,8 +28,8 @@ function [averages, radii] = angularAverage(im)
 %
 %     N = [1024,1024];
 %     fresnel = 1e-2;
-%     [xi_1,xi_2] = fftfreq(N, 1, true);
-%     ctf = fftshift(sin((xi_1.^2 + xi_2.^2)/(4*pi*fresnel)));
+%     xi2 = fftfreqNormSq(N, 1);
+%     ctf = fftshift(sin(xi2/(4*pi*fresnel)));
 %     [ctfAveraged, radii] = angularAverage(ctf);
 %     
 %     figure('name', 'image'); imagesc(ctf); colorbar;

functions/imageProcessing/cropPadWindow/padToSize.m

@@ -76,9 +76,9 @@ end
 padsize = floor((outputSize - imageSize)/2);
 padsize2 = outputSize - (imageSize + 2 * padsize);
 
-result = padarray(im, padsize, 'both', padval);
+result = padarray(im, padsize, padval, 'both');
 if any(padsize2 > 0)
-    result = padarray(result, padsize2, 'pre', padval);
+    result = padarray(result, padsize2, padval, 'pre');
 end
 
 end

functions/phaseRetrieval/holographic/phaserec_holotie.m

@@ -154,10 +154,10 @@ laplaceKernel = -fftfreqNormSq([ny, nx]);
 switch (settings.reg_type)
     case 'none'
         % avoid division by 0
-        laplaceKernel(0,0) = 1;
+        laplaceKernel(1,1) = 1;
         invLaplaceFilt = 1 ./ laplaceKernel;
         % set zero frequency (singularity) to 0
-        invLaplaceFilt(0,0) = 0;
+        invLaplaceFilt(1,1) = 0;
     case 'mba'
         invLaplaceFilt = -1 ./ (-laplaceKernel + settings.reg_alpha);
     case 'tikhonov'

functions/phaseRetrieval/holographic/private/ctfRegWeights.m

@@ -29,7 +29,7 @@ end
 
 ndim = numel(sizeHolo);
 xi = cell([ndim,1]);
-[xi{:}] = fftfreq(sizeHolo, 1, false);
+[xi{:}] = fftfreq(sizeHolo, 1);
 XiSqDivFresnel = 0;
 for jj = 1:ndim
     XiSqDivFresnel = XiSqDivFresnel + gpuArrayIf(xi{jj}, gpuFlag).^2 ./ (2*pi^2*fresnelNumbers(jj));

functions/tomography/ringremove.m

@@ -6,8 +6,7 @@ function sinoCorrected = ringremove(sino,settings)
 % Removes ring artifacts in tomographic reconstructions by reducing the amount of
 % horizontal lines in the given sinogram. Ring removal is either based on the method
 % proposed by Ketcham et al. :cite:`Ketcham_PS_2006`, denoted as 'additive', or on the method proposed
-% by Muench et al. :cite:`Muench_OE_2009`, denoted as 'wavelet'. Large parts of the code for the
-% wavelet-based approach are taken from the original paper
+% by Muench et al. :cite:`Muench_OE_2009`, denoted as 'wavelet'.
 %
 % Parameters
 % ----------
@@ -38,7 +37,7 @@ function sinoCorrected = ringremove(sino,settings)
 %     Highest decomposition level in the wavelet decomposition when using method
 %     'wavelet'.
 % wavelet_Sigma : Default = 1
-%     Standard deviation for the Gaussian filtering of the horizontal part of the
+%     Width of the Gaussian filter kernel for the horizontal part of the
 %     sinogram in Fourier space when using method 'wavelet'.
 %
 % Returns
@@ -123,7 +122,6 @@ settings = completeStruct(settings, defaults);
 
 switch settings.method
     case 'additive'
-        numAngles = size(sino,2);
         % averaging of sinogram along the angle
         sinoMean = mean(sino,2);
         
@@ -140,50 +138,43 @@ switch settings.method
         end
         % identification of horizontal stripes
         correctionProfile = sinoMean - sinoMeanSmoothed;
-        % creation of 2d correction mask
-        correction2D = repmat(correctionProfile,[1 numAngles]);
         % sinogram correction
-        sinoCorrected = sino - settings.additive_Strength*correction2D;
+        sinoCorrected = sino - settings.additive_Strength * correctionProfile;
        
     case 'wavelet'
-        numAngles = size(sino,2);
-        sizeProjs = size(sino,1);
-        
-        % wavelet decomposition
-        for decompIdx = 1:settings.wavelet_DecNum
-            [sino,horzPart{decompIdx},vertPart{decompIdx},diagPart{decompIdx}] = ...
-                dwt2(sino,settings.wavelet_Wname);
+		
+        % perform multilevel wavelet decomposition		
+		H = cell(settings.wavelet_DecNum, 1);
+		V = cell(settings.wavelet_DecNum, 1);
+		D = cell(settings.wavelet_DecNum, 1);
+		S = cell(settings.wavelet_DecNum, 1);
+				
+        for l = 1:settings.wavelet_DecNum
+			S{l} = size(sino);
+            [sino, H{l}, V{l}, D{l}] = dwt2(sino,settings.wavelet_Wname);
         end
         
-        % FFT transform of horizontal frequency bands
-        for decompIdx = 1:settings.wavelet_DecNum
+        % filter horizontal frequency bands
+        for l = 1:settings.wavelet_DecNum
             % FFT
-            horzFourier = fftshift(fft(horzPart{decompIdx}'));
-            [my,mx] = size(horzFourier);
+            horzFourier = fft(H{l}');
             
             % damping of horizontal stripe information
-            dampGauss = 1 - exp(-(-floor(my/2):-floor(my/2) + my - 1).^2/(2*settings.wavelet_Sigma^2));
-            horzFourier = horzFourier.*repmat(dampGauss',1,mx);
+			m = size(horzFourier,1);
+            k = fftfreq(m, 2*pi/m);
+            dampGauss = 1 - exp(-1/2 * k.^2 / settings.wavelet_Sigma^2 );
+            horzFourier = horzFourier .* dampGauss;
             
             % inverse FFT
-            horzPart{decompIdx} = ifft(ifftshift(horzFourier))';
-        end
-        
-        % wavelet reconstruction
-        sinoCorrected = sino;
-        for decompIdx = settings.wavelet_DecNum:-1:1
-            sinoCorrected = sinoCorrected(1:size(vertPart{decompIdx},1),1:size(vertPart{decompIdx},2));
-            sinoCorrected = idwt2(sinoCorrected,horzPart{decompIdx},vertPart{decompIdx},...
-                diagPart{decompIdx},settings.wavelet_Wname);
+            H{l} = ifft(horzFourier)';
         end
         
-        % make sure that corrected sinogram has the same size as the original sinogram
-        if size(sinoCorrected,2) == numAngles + 1
-            sinoCorrected = sinoCorrected(:,1:end - 1);
-        end
-        if size(sinoCorrected,1) == sizeProjs + 1
-            sinoCorrected = sinoCorrected(1:end - 1,:);
-        end
+		% wavelet reconstruction
+		sinoCorrected = sino;
+		for l = settings.wavelet_DecNum:-1:1
+			sinoCorrected = sinoCorrected(1:size(V{l},1),1:size(V{l},2));
+			sinoCorrected = idwt2(sinoCorrected,H{l}, V{l}, D{l},settings.wavelet_Wname, S{l});
+		end	
         
     otherwise
         error('Choose between ringremove based on "additive" or "wavelet" correction for setting "method"');