Merge pull request #65 from elgw/dev

v 0.4.2

CHANGELOG.md

@@ -1,5 +1,10 @@
 # CHANGELOG
 
+## 0.4.2
+- VkFFT upgraded from v1.3.3 to v1.3.4.
+- Fixed an issue preventing anything than `--bq 0` to be used on
+  Windows (one array was not properly initialized to 0).
+
 ## 0.4.1
 - Fixes the GPU memory leak reported in [issue
   #62](https://github.com/elgw/deconwolf/issues/62).
@@ -176,19 +181,19 @@ Results when testing on the microtubules image shown below.
    commenting in/out the corresponding lines in the makefile.
  - Cleaned up the output of `dw --version`
 
-## v 0.2.7
+## 0.2.7
 - Converted a few minor code paths to execute in parallel by OpenMP
   directives.
 
-## v 0.2.6
+## 0.2.6
 - Using ISO 8601 in log files, e.g., `2023-02-14T11:14:14`.
 
-## v 0.2.5
+## 0.2.5
 - Added the **--xyz** option to **dw maxproj**, for creating max
   projections along the three axes and collecting them on a single 2D
   image.
 
-## v 0.2.4
+## 0.2.4
 - **dw --help** now shows the additional commands/modules available.
 - Reading 16-bit tif files with **TIFFReadEncodedStrip** instead of
   **TIFFReadRawStrip**. Some programs saves tiff files in other ways
@@ -199,26 +204,26 @@ Results when testing on the microtubules image shown below.
 - Setting the background level automatically to min(image) unless
   specified with **--bg**.
 
-## v 0.2.3
+## 0.2.3
 - Fixed some errors introduced in v 0.2.2, especially the **dw
   maxproj** was broken.
 - added the subcommand **dw merge**. To be used to merge single z-planes
   into a 3D volume.
 
-## v 0.2.2
+## 0.2.2
 - Can deconvolve using clFFT, when compiled with **OPENCL=1** two new
   methods appear, **--method shbcl** and **--shbcl2**, the first using
   clFFT only for the Fourier transforms, the latter using OpenCL for
   the whole deconvolution procedure. Uses quite much GPU memory which
   is something to improve upon in future version, possibly by
   switching to vkFFT.
 
-## v 0.1.1
+## 0.1.1
 - Added experimental **dw imshift** for shifting images, also shift
   estimation using normalized cross correlation with **dw imshift
   --ref file.tif**. Might be extended to basic tiling etc.
 
-## v 0.1.0
+## 0.1.0
 - Implements the 'Scaled Heavy Ball'. More
 memory efficient than eve and about the same speed and image
 quality. Might become the default method.
@@ -238,20 +243,20 @@ iteration to a separate tsv file for easier plotting and analysis.
 - Three different stopping criteria: Relative error (default) Fixed
   number of iterations or at an absolute error.
 
-## v 0.0.26
+## 0.0.26
 - **dw maxproj** works with file that are not in the current folder.
 - Fixed **--iterdump** not always working.
 
-## v. 0.0.25
+## 0.0.25
 - Builds with cuFFT on Linux, use `make CUFFT=1 -B`, requires a CUDA
 compatible GPU and of course the cuFFT library installed.
 
-## v. 0.0.24
+## 0.0.24
 - Tested on CentOS, install both with make and meson.
 - Fixed a memory leak with the **--tilesize** option causing
 crashed sometimes.
 
-## v. 0.0.23
+## 0.0.23
 - Added 'meson.build' files in order for deconwolf to be built by
 [The Meson Build system](https://mesonbuild.com/), tested to work
 on both Ubuntu 21.10 and MacOS (on x86_64 hardware).
@@ -263,17 +268,17 @@ document.
 - Aborting if the number of threads is set < 1.
 - The algorithm is still unchanged since v 0.0.20.
 
-## v. 0.0.22
+## 0.0.22
 - Fixed double free-bug in tiling mode.
 
-## v. 0.0.21
+## 0.0.21
 - Updated documentation and man-pages based on markdown files
 for easier updating.
 - Provides `makefile-freebsd` for building on FreeBSD 13.0
 - Changed behavior when too few input arguments are given to
 only give a two-line message.
 
-## v. 0.0.20
+## 0.0.20
 - Changing acceleration technique to use
 'Exponential Vector Extrapolation' (EVE) described in Biggs PhD thesis.
 Deconvolved images get higher MSE but much lower I-div.
@@ -291,13 +296,13 @@ however that is not recommended.
 experimental option. 14 percent faster on a small test image, varied
 results on larger images.
 
-## v. 0.0.19
+## 0.0.19
 - Using lanczos5 instead of lanczos3 for the PSF generation. As a result
 GSL_EROUND is not raised for the test cases.
 - Faster PSF generation, using more symmetries.
 - dw_bw can now use more than one thread (wrongly disabled in v 0.0.18).
 
-## v. 0.0.18
+## 0.0.18
 - Provided install instructions for Windows 10.
 - Fixed some mismatching fftwf_malloc/fftwf_free where they were
 mixed up with malloc/free causing crashes on Windows.
@@ -306,7 +311,7 @@ building with cmake. It is also possible to cross compile for Windows
 on Linux although it takes some effort to collect the DLL files for the
 dependencies.
 
-## v. 0.0.17
+## 0.0.17
 - Fixed some bugs in the PSF generation code that did affect the accuracy
 of the pixels in the PSF.
 - Stared to use GSL for numerical integration. It remains to change the

INSTALL.md

@@ -118,7 +118,6 @@ The dependencies can get retrieved by vcpkg:
 ``` shell
 git clone https://github.com/microsoft/vcpkg
 .\vcpkg\bootstrap-vcpkg.bat
-.\vcpkg integrate install
 .\vcpkg\vcpkg.exe install fftw3[threads]
 .\vcpkg\vcpkg.exe install tiff
 .\vcpkg\vcpkg.exe install gsl
@@ -138,9 +137,17 @@ cmake "-DCMAKE_TOOLCHAIN_FILE=C:/YOUR/OWN/PATH/vcpkg/scripts/buildsystems/vcpkg.
 ```
 Important: Please use the correct path to `vcpkg.cmake`.
 
-Open visual studio and compile it from there. Don't forget to change the build type to release. 
+Open the visual studio solution and oo to Linker->Input->Additional
+dependencies and add:
 
-If you are a windows developer and reading this, please help us out get make this process smoother (if possible). 
+``` shell
+libomp.lib
+```
+
+Change the build type from debug to release and compile.
+
+If you are a windows developer and reading this, please help us out
+to make the build process smoother!
 
 ## FreeBSD
 - Use `gmake`, not `make`.

README.md

@@ -1,4 +1,4 @@
-# deconwolf v0.4.1
+# deconwolf v0.4.2
 
 **deconwolf**[^9] is a software for 3-D deconvolution of fluorescent wide-field
 images:

src/VkFFT/vkFFT/vkFFT.h

@@ -107,6 +107,6 @@
 #include "vkFFT/vkFFT_AppManagement/vkFFT_DeleteApp.h"
 
 static inline int VkFFTGetVersion() {
-	return 10303; //X.XX.XX format
+	return 10304; //X.XX.XX format
 }
 #endif

src/dw_nuclei.c

@@ -89,9 +89,9 @@ static void opts_free(opts * s)
  * than green set to 2. Everything else set to 0.
  * http://www.libpng.org/pub/png/libpng-manual.txt
  */
-fim_t * fim_png_read_green_red(char * fname)
+fimo * fim_png_read_green_red(char * fname)
 {
-    fim_t * F = NULL;
+    fimo * F = NULL;
 
     png_image image;
     memset(&image, 0, sizeof(image));
@@ -113,7 +113,7 @@ fim_t * fim_png_read_green_red(char * fname)
             size_t N = image.height;
             //printf("M=%zu, N=%zu image buffer %u b\n", M, N, PNG_IMAGE_SIZE(image));
 
-            F = malloc(sizeof(fim_t));
+            F = malloc(sizeof(fimo));
             assert(F != NULL);
             F->M = M;
             F->N = N;
@@ -291,7 +291,7 @@ static int float_arg_max(const float * v, size_t N)
     return argmax;
 }
 
-fim_t * get_reduction(opts * s, char * file)
+fimo * get_reduction(opts * s, char * file)
 {
 
 
@@ -325,7 +325,7 @@ fim_t * get_reduction(opts * s, char * file)
     {
         float * maxI = fim_maxproj(I, M, N, P);
         free(I);
-        fim_t * result = fim_image_from_array(maxI, M, N, 1);
+        fimo * result = fim_image_from_array(maxI, M, N, 1);
         free(maxI);
         return result;
     }
@@ -336,20 +336,20 @@ fim_t * get_reduction(opts * s, char * file)
         {
             printf("Finding focus\n"); fflush(stdout);
         }
-        fim_t * II = fim_image_from_array(I, M, N, P);
+        fimo * II = fim_image_from_array(I, M, N, P);
         free(I);
         float sigma = 1;
         float * gm = fim_focus_gm(II, sigma);
         int slice = float_arg_max(gm, II->P);
-        fim_t * result = malloc(sizeof(fim_t));
+        fimo * result = malloc(sizeof(fimo));
         assert(result != NULL);
         result->M = M;
         result->N = N;
         result->P = 1;
         result->V = malloc(M*N*sizeof(float));
         assert(result->V != NULL);
         memcpy(result->V, II->V+slice*M*N, M*N*sizeof(float));
-        fimt_free(II);
+        fimo_free(II);
         if(s->verbose > 0)
         {
             printf("Returning slice %d\n", slice); fflush(stdout);
@@ -387,7 +387,7 @@ void segment_image_rf(opts * s, PrfForest * F, char * file)
     printf("\n");
 
     /* Read the image */
-    fim_t * redu = get_reduction(s, file);
+    fimo * redu = get_reduction(s, file);
     assert(redu != NULL);
     printf("1\n"); fflush(stdout);
     /* Extract features */
@@ -453,7 +453,7 @@ void segment_image_rf(opts * s, PrfForest * F, char * file)
 
     fim_tiff_write_noscale(outfile, result, NULL,
                            redu->M, redu->N, 1);
-    fimt_free(redu);
+    fimo_free(redu);
     return;
 }
 
@@ -471,7 +471,7 @@ PrfForest * loop_training_data(opts * s, float * features_cm,
         }
 
         /* Read annotated image */
-        fim_t * anno = fim_png_read_green_red(s->anno_label);
+        fimo * anno = fim_png_read_green_red(s->anno_label);
         assert(anno != NULL);
 
         /* Append one columns for the annotations */
@@ -551,7 +551,7 @@ PrfForest * loop_training_data(opts * s, float * features_cm,
         }
         free(features_cma);
         free(features_cma_train);
-        fimt_free(anno);
+        fimo_free(anno);
         free(result);
     }
 
@@ -562,7 +562,7 @@ void random_forest_pipeline(opts * s, int argc, char ** argv)
 {
 
     /* Read raw image */
-    fim_t * anno_raw = get_reduction(s, s->anno_image);
+    fimo * anno_raw = get_reduction(s, s->anno_image);
 
     /* Extract features */
     ftab_t * features = fim_features_2d(anno_raw);

src/dw_psf.c

@@ -545,14 +545,14 @@ static void downsample_integrate(float * A, float * B, int nA, int nB)
     free(K);
 }
 
-fim_t * gen_psf(opts * s, double lambda)
+fimo * gen_psf(opts * s, double lambda)
     {
     if(s->verbose > 2) { printf("Allocating for output\n"); fflush(stdout); }
 
     double optical_dx = s->optical.dx/s->oversampling;
 
     fim_tiff_init();
-    fim_t * PSF = fimt_zeros(s->M*s->oversampling, s->M*s->oversampling, s->P);
+    fimo * PSF = fimo_zeros(s->M*s->oversampling, s->M*s->oversampling, s->P);
 
     double Fs = 1/optical_dx;
     double Fn = Fs/2;
@@ -765,7 +765,7 @@ static float * conv2d_float(float * A, float * B,
 }
 
 
-static void pinhole_convolution(opts * s, fim_t * PSF)
+static void pinhole_convolution(opts * s, fimo * PSF)
 {
     /* Convolve PSF by a square filter. The specified pinhole is the
        diameter. */
@@ -840,7 +840,7 @@ static void dw_psf(opts * s)
         printf("Calculating emission PSF\n");
     }
 
-    fim_t * PSF = gen_psf(s, s->optical.lambda);
+    fimo * PSF = gen_psf(s, s->optical.lambda);
     if(s->optical.lambda2 != 0)
     {
         /* We will generate a PSF for a confocal microscope with
@@ -855,22 +855,22 @@ static void dw_psf(opts * s)
         {
             printf("Calculating excitation PSF\n");
         }
-        fim_t * PSF2 = gen_psf(s, s->optical.lambda2);
+        fimo * PSF2 = gen_psf(s, s->optical.lambda2);
         //printf("Multiplying PSFs\n");
-        for(size_t kk = 0; kk<fimt_nel(PSF); kk++)
+        for(size_t kk = 0; kk<fimo_nel(PSF); kk++)
         {
             PSF->V[kk] *= PSF2->V[kk];
         }
-        fimt_free(PSF2);
+        fimo_free(PSF2);
     }
 
 
     if(s->verbose > 0)
     {
         printf("Writing to %s\n", s->outfile);
     }
-    float sum = fimt_sum(PSF);
-    fim_mult_scalar(PSF->V, fimt_nel(PSF), 1.0/sum);
+    float sum = fimo_sum(PSF);
+    fim_mult_scalar(PSF->V, fimo_nel(PSF), 1.0/sum);
 
     ttags * T = ttags_new();
     char * swstring = malloc(1024);
@@ -886,7 +886,7 @@ static void dw_psf(opts * s)
                          T,
                          PSF->M, PSF->N, PSF->P);
     ttags_free(&T);
-    fimt_free(PSF);
+    fimo_free(PSF);
 }
 
 int dw_psf_cli(int argc, char ** argv)

src/dw_psf_sted.c

@@ -289,9 +289,9 @@ static void argparsing(int argc, char ** argv, opts * s)
 
 
 
-static fim_t * gen_psf(opts * s)
+static fimo * gen_psf(opts * s)
 {
-    fim_t * PSF = fimt_zeros(s->M, s->M, s->P);
+    fimo * PSF = fimo_zeros(s->M, s->M, s->P);
     const double gamma = s->gamma;
     const double sigma = s->sigma;
 
@@ -357,10 +357,10 @@ static void dw_psf_sted(opts * s)
     {
         printf("Writing to %s\n", s->outfile);
     }
-    fim_t * PSF = gen_psf(s);
+    fimo * PSF = gen_psf(s);
 
-    float sum = fimt_sum(PSF);
-    fim_mult_scalar(PSF->V, fimt_nel(PSF), 1.0/sum);
+    float sum = fimo_sum(PSF);
+    fim_mult_scalar(PSF->V, fimo_nel(PSF), 1.0/sum);
 
     ttags * T = ttags_new();
     char * swstring = malloc(1024);
@@ -377,7 +377,7 @@ static void dw_psf_sted(opts * s)
                          T,
                          PSF->M, PSF->N, PSF->P);
     ttags_free(&T);
-    fimt_free(PSF);
+    fimo_free(PSF);
 }
 
 int dw_psf_sted_cli(int argc, char ** argv)