Merge pull request #52 from vais-ral/NonlDiffusion

Nonlinear diffusion module

14 files changed, 646 insertions, 24 deletions

diff --git a/Wrappers/Matlab/demos/demoMatlab_3Ddenoise.m b/Wrappers/Matlab/demos/demoMatlab_3Ddenoise.m
index fb55097..973d060 100644
--- a/Wrappers/Matlab/demos/demoMatlab_3Ddenoise.m
+++ b/Wrappers/Matlab/demos/demoMatlab_3Ddenoise.m
@@ -53,6 +53,23 @@ figure; imshow(u_sb(:,:,15), [0 1]); title('SB-TV denoised volume (CPU)');
 % tic; u_sbG = SB_TV_GPU(single(vol3D), lambda_reg, iter_sb, epsil_tol); toc; 
 % figure; imshow(u_sbG(:,:,15), [0 1]); title('SB-TV denoised volume (GPU)');
 %%
+%%
+fprintf('Denoise a volume using Nonlinear-Diffusion model (CPU) \n');
+iter_diff = 300; % number of diffusion iterations
+lambda_regDiff = 0.06; % regularisation for the diffusivity 
+sigmaPar = 0.04; % edge-preserving parameter
+tau_param = 0.025; % time-marching constant 
+tic; u_diff = NonlDiff(single(vol3D), lambda_regDiff, sigmaPar, iter_diff, tau_param, 'Huber'); toc; 
+figure; imshow(u_diff(:,:,15), [0 1]); title('Diffusion denoised volume (CPU)');
+%%
+% fprintf('Denoise a volume using Nonlinear-Diffusion model (GPU) \n');
+% iter_diff = 300; % number of diffusion iterations
+% lambda_regDiff = 0.06; % regularisation for the diffusivity 
+% sigmaPar = 0.04; % edge-preserving parameter
+% tau_param = 0.025; % time-marching constant 
+% tic; u_diff_g = NonlDiff_GPU(single(vol3D), lambda_regDiff, sigmaPar, iter_diff, tau_param, 'Huber'); toc; 
+% figure; imshow(u_diff_g(:,:,15), [0 1]); title('Diffusion denoised volume (GPU)');
+%%
 
 %>>>>>>>>>>>>>> MULTI-CHANNEL priors <<<<<<<<<<<<<<< %
 fprintf('Denoise a volume using the FGP-dTV model (CPU) \n');
diff --git a/Wrappers/Matlab/demos/demoMatlab_denoise.m b/Wrappers/Matlab/demos/demoMatlab_denoise.m
index dab98dc..4a0a19a 100644
--- a/Wrappers/Matlab/demos/demoMatlab_denoise.m
+++ b/Wrappers/Matlab/demos/demoMatlab_denoise.m
@@ -46,6 +46,22 @@ figure; imshow(u_sb, [0 1]); title('SB-TV denoised image (CPU)');
 % tic; u_sbG = SB_TV_GPU(single(u0), lambda_reg, iter_sb, epsil_tol); toc; 
 % figure; imshow(u_sbG, [0 1]); title('SB-TV denoised image (GPU)');
 %%
+fprintf('Denoise using Nonlinear-Diffusion model (CPU) \n');
+iter_diff = 800; % number of diffusion iterations
+lambda_regDiff = 0.06; % regularisation for the diffusivity 
+sigmaPar = 0.04; % edge-preserving parameter
+tau_param = 0.025; % time-marching constant 
+tic; u_diff = NonlDiff(single(u0), lambda_regDiff, sigmaPar, iter_diff, tau_param, 'Huber'); toc; 
+figure; imshow(u_diff, [0 1]); title('Diffusion denoised image (CPU)');
+%%
+% fprintf('Denoise using Nonlinear-Diffusion model (GPU) \n');
+% iter_diff = 800; % number of diffusion iterations
+% lambda_regDiff = 0.06; % regularisation for the diffusivity 
+% sigmaPar = 0.04; % edge-preserving parameter
+% tau_param = 0.025; % time-marching constant 
+% tic; u_diff_g = NonlDiff_GPU(single(u0), lambda_regDiff, sigmaPar, iter_diff, tau_param, 'Huber'); toc; 
+% figure; imshow(u_diff_g, [0 1]); title('Diffusion denoised image (GPU)');
+%%
 %>>>>>>>>>>>>>> MULTI-CHANNEL priors <<<<<<<<<<<<<<< %
 
 fprintf('Denoise using the FGP-dTV model (CPU) \n');
diff --git a/Wrappers/Matlab/mex_compile/compileCPU_mex.m b/Wrappers/Matlab/mex_compile/compileCPU_mex.m
index 9892d73..ec799bd 100644
--- a/Wrappers/Matlab/mex_compile/compileCPU_mex.m
+++ b/Wrappers/Matlab/mex_compile/compileCPU_mex.m
@@ -20,7 +20,10 @@ movefile FGP_dTV.mex* ../installed/
 mex TNV.c TNV_core.c utils.c CFLAGS="\$CFLAGS -fopenmp -Wall -std=c99" LDFLAGS="\$LDFLAGS -fopenmp"
 movefile TNV.mex* ../installed/
 
-delete SB_TV_core* ROF_TV_core* FGP_TV_core* FGP_dTV_core* TNV_core* utils* CCPiDefines.h
+mex NonlDiff.c Diffusion_core.c utils.c CFLAGS="\$CFLAGS -fopenmp -Wall -std=c99" LDFLAGS="\$LDFLAGS -fopenmp"
+movefile NonlDiff.mex* ../installed/
+
+delete SB_TV_core* ROF_TV_core* FGP_TV_core* FGP_dTV_core* TNV_core* utils* Diffusion_core* CCPiDefines.h
 
 fprintf('%s \n', 'All successfully compiled!');
 
diff --git a/Wrappers/Matlab/mex_compile/compileGPU_mex.m b/Wrappers/Matlab/mex_compile/compileGPU_mex.m
index 3dbeb8a..55b51eb 100644
--- a/Wrappers/Matlab/mex_compile/compileGPU_mex.m
+++ b/Wrappers/Matlab/mex_compile/compileGPU_mex.m
@@ -31,7 +31,11 @@ movefile SB_TV_GPU.mex* ../installed/
 mex -g -I/usr/local/cuda-7.5/include -L/usr/local/cuda-7.5/lib64 -lcudart -lcufft -lmwgpu FGP_dTV_GPU.cpp dTV_FGP_GPU_core.o
 movefile FGP_dTV_GPU.mex* ../installed/
 
-delete TV_ROF_GPU_core* TV_FGP_GPU_core* TV_SB_GPU_core* dTV_FGP_GPU_core* CCPiDefines.h
+!/usr/local/cuda/bin/nvcc -O0 -c NonlDiff_GPU_core.cu -Xcompiler -fPIC -I~/SOFT/MATLAB9/extern/include/
+mex -g -I/usr/local/cuda-7.5/include -L/usr/local/cuda-7.5/lib64 -lcudart -lcufft -lmwgpu NonlDiff_GPU.cpp NonlDiff_GPU_core.o
+movefile NonlDiff_GPU.mex* ../installed/
+
+delete TV_ROF_GPU_core* TV_FGP_GPU_core* TV_SB_GPU_core* dTV_FGP_GPU_core* NonlDiff_GPU_core* CCPiDefines.h
 fprintf('%s \n', 'All successfully compiled!');
 
 cd ../../
diff --git a/Wrappers/Matlab/mex_compile/regularisers_CPU/NonlDiff.c b/Wrappers/Matlab/mex_compile/regularisers_CPU/NonlDiff.c
new file mode 100644
index 0000000..e05f5d4
--- /dev/null
+++ b/Wrappers/Matlab/mex_compile/regularisers_CPU/NonlDiff.c
@@ -0,0 +1,87 @@
+/*
+ * This work is part of the Core Imaging Library developed by
+ * Visual Analytics and Imaging System Group of the Science Technology
+ * Facilities Council, STFC
+ *
+ * Copyright 2017 Daniil Kazantsev
+ * Copyright 2017 Srikanth Nagella, Edoardo Pasca
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ * http://www.apache.org/licenses/LICENSE-2.0
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include "matrix.h"
+#include "mex.h"
+#include "Diffusion_core.h"
+
+/* C-OMP implementation of linear and nonlinear diffusion with the regularisation model [1] (2D/3D case)
+ * The minimisation is performed using explicit scheme.
+ *
+ * Input Parameters:
+ * 1. Noisy image/volume
+ * 2. lambda - regularization parameter
+ * 3. Edge-preserving parameter (sigma), when sigma equals to zero nonlinear diffusion -> linear diffusion
+ * 4. Number of iterations, for explicit scheme >= 150 is recommended  [OPTIONAL parameter]
+ * 5. tau - time-marching step for explicit scheme [OPTIONAL parameter]
+ * 6. Penalty type: 1 - Huber, 2 - Perona-Malik, 3 - Tukey Biweight [OPTIONAL parameter]
+ *
+ * Output:
+ * [1] Regularized image/volume
+ *
+ * This function is based on the paper by
+ * [1] Perona, P. and Malik, J., 1990. Scale-space and edge detection using anisotropic diffusion. IEEE Transactions on pattern analysis and machine intelligence, 12(7), pp.629-639.
+ */
+
+void mexFunction(
+        int nlhs, mxArray *plhs[],
+        int nrhs, const mxArray *prhs[])
+        
+{
+    int number_of_dims, iter_numb, dimX, dimY, dimZ, penaltytype;
+    const int  *dim_array;
+    float *Input, *Output=NULL, lambda, tau, sigma;
+    
+    dim_array = mxGetDimensions(prhs[0]);
+    number_of_dims = mxGetNumberOfDimensions(prhs[0]);
+    
+    /*Handling Matlab input data*/
+    Input  = (float *) mxGetData(prhs[0]);
+    lambda =  (float) mxGetScalar(prhs[1]); /* regularization parameter */
+    sigma = (float) mxGetScalar(prhs[2]); /* Edge-preserving parameter */
+    iter_numb = 300; /* iterations number */
+    tau = 0.025; /* marching step parameter */
+    penaltytype = 1; /* Huber penalty by default */
+    
+    if (mxGetClassID(prhs[0]) != mxSINGLE_CLASS) {mexErrMsgTxt("The input image must be in a single precision"); }
+    if ((nrhs < 3) || (nrhs > 6)) mexErrMsgTxt("At least 3 parameters is required, all parameters are: Image(2D/3D), Regularisation parameter, Edge-preserving parameter, iterations number, time-marching constant, penalty type - Huber, PM or Tukey");
+    if ((nrhs == 4) || (nrhs == 5) || (nrhs == 6))  iter_numb = (int) mxGetScalar(prhs[3]); /* iterations number */
+    if ((nrhs == 5) || (nrhs == 6))  tau =  (float) mxGetScalar(prhs[4]); /* marching step parameter */
+    if (nrhs == 6)  {
+        char *penalty_type;
+        penalty_type = mxArrayToString(prhs[5]); /* Huber, PM or Tukey 'Huber' is the default */
+        if ((strcmp(penalty_type, "Huber") != 0) && (strcmp(penalty_type, "PM") != 0) && (strcmp(penalty_type, "Tukey") != 0)) mexErrMsgTxt("Choose penalty: 'Huber', 'PM' or 'Tukey',");
+        if (strcmp(penalty_type, "Huber") == 0)  penaltytype = 1;  /* enable 'Huber' penalty */
+        if (strcmp(penalty_type, "PM") == 0)  penaltytype = 2;  /* enable Perona-Malik penalty */
+        if (strcmp(penalty_type, "Tukey") == 0)  penaltytype = 3;  /* enable Tikey Biweight penalty */
+        mxFree(penalty_type);
+    }    
+    
+    /*Handling Matlab output data*/
+    dimX = dim_array[0]; dimY = dim_array[1]; dimZ = dim_array[2];
+    
+    /* output arrays*/
+    if (number_of_dims == 2) {
+        dimZ = 1; /*2D case*/
+        /* output image/volume */
+        Output = (float*)mxGetPr(plhs[0] = mxCreateNumericArray(2, dim_array, mxSINGLE_CLASS, mxREAL));
+    }
+    if (number_of_dims == 3) Output = (float*)mxGetPr(plhs[0] = mxCreateNumericArray(3, dim_array, mxSINGLE_CLASS, mxREAL));
+    
+    Diffusion_CPU_main(Input, Output, lambda, sigma, iter_numb, tau, penaltytype, dimX, dimY, dimZ);
+}
\ No newline at end of file
diff --git a/Wrappers/Matlab/mex_compile/regularisers_GPU/NonlDiff_GPU.cpp b/Wrappers/Matlab/mex_compile/regularisers_GPU/NonlDiff_GPU.cpp
new file mode 100644
index 0000000..bfba9ea
--- /dev/null
+++ b/Wrappers/Matlab/mex_compile/regularisers_GPU/NonlDiff_GPU.cpp
@@ -0,0 +1,90 @@
+/*
+ * This work is part of the Core Imaging Library developed by
+ * Visual Analytics and Imaging System Group of the Science Technology
+ * Facilities Council, STFC
+ *
+ * Copyright 2017 Daniil Kazantsev
+ * Copyright 2017 Srikanth Nagella, Edoardo Pasca
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ * http://www.apache.org/licenses/LICENSE-2.0
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include "matrix.h"
+#include "mex.h"
+#include <stdio.h>
+#include <string.h>
+#include "NonlDiff_GPU_core.h"
+
+/* CUDA implementation of linear and nonlinear diffusion with the regularisation model [1,2] (2D/3D case)
+ * The minimisation is performed using explicit scheme. 
+ *
+ * Input Parameters:
+ * 1. Noisy image/volume 
+ * 2. lambda - regularization parameter
+ * 3. Edge-preserving parameter (sigma), when sigma equals to zero nonlinear diffusion -> linear diffusion
+ * 4. Number of iterations, for explicit scheme >= 150 is recommended 
+ * 5. tau - time-marching step for explicit scheme
+ * 6. Penalty type: 1 - Huber, 2 - Perona-Malik, 3 - Tukey Biweight
+ *
+ * Output:
+ * [1] Regularized image/volume 
+ *
+ * This function is based on the paper by
+ * [1] Perona, P. and Malik, J., 1990. Scale-space and edge detection using anisotropic diffusion. IEEE Transactions on pattern analysis and machine intelligence, 12(7), pp.629-639.
+ * [2] Black, M.J., Sapiro, G., Marimont, D.H. and Heeger, D., 1998. Robust anisotropic diffusion. IEEE Transactions on image processing, 7(3), pp.421-432.
+ */
+
+void mexFunction(
+        int nlhs, mxArray *plhs[],
+        int nrhs, const mxArray *prhs[])
+        
+{
+    int number_of_dims, iter_numb, dimX, dimY, dimZ, penaltytype;
+    const int  *dim_array;
+    float *Input, *Output=NULL, lambda, tau, sigma;
+    
+    dim_array = mxGetDimensions(prhs[0]);
+    number_of_dims = mxGetNumberOfDimensions(prhs[0]);
+    
+    /*Handling Matlab input data*/
+    Input  = (float *) mxGetData(prhs[0]);
+    lambda =  (float) mxGetScalar(prhs[1]); /* regularization parameter */
+    sigma = (float) mxGetScalar(prhs[2]); /* Edge-preserving parameter */
+    iter_numb = 300; /* iterations number */
+    tau = 0.025; /* marching step parameter */
+    penaltytype = 1; /* Huber penalty by default */
+    
+    if (mxGetClassID(prhs[0]) != mxSINGLE_CLASS) {mexErrMsgTxt("The input image must be in a single precision"); }
+    if ((nrhs < 3) || (nrhs > 6)) mexErrMsgTxt("At least 3 parameters is required, all parameters are: Image(2D/3D), Regularisation parameter, Edge-preserving parameter, iterations number, time-marching constant, penalty type - Huber, PM or Tukey");
+    if ((nrhs == 4) || (nrhs == 5) || (nrhs == 6))  iter_numb = (int) mxGetScalar(prhs[3]); /* iterations number */
+    if ((nrhs == 5) || (nrhs == 6))  tau =  (float) mxGetScalar(prhs[4]); /* marching step parameter */
+    if (nrhs == 6)  {
+        char *penalty_type;
+        penalty_type = mxArrayToString(prhs[5]); /* Huber, PM or Tukey 'Huber' is the default */
+        if ((strcmp(penalty_type, "Huber") != 0) && (strcmp(penalty_type, "PM") != 0) && (strcmp(penalty_type, "Tukey") != 0)) mexErrMsgTxt("Choose penalty: 'Huber', 'PM' or 'Tukey',");
+        if (strcmp(penalty_type, "Huber") == 0)  penaltytype = 1;  /* enable 'Huber' penalty */
+        if (strcmp(penalty_type, "PM") == 0)  penaltytype = 2;  /* enable Perona-Malik penalty */
+        if (strcmp(penalty_type, "Tukey") == 0)  penaltytype = 3;  /* enable Tikey Biweight penalty */
+        mxFree(penalty_type);
+    }    
+    
+    /*Handling Matlab output data*/
+    dimX = dim_array[0]; dimY = dim_array[1]; dimZ = dim_array[2];
+    
+    /* output arrays*/
+    if (number_of_dims == 2) {
+        dimZ = 1; /*2D case*/
+        /* output image/volume */
+        Output = (float*)mxGetPr(plhs[0] = mxCreateNumericArray(2, dim_array, mxSINGLE_CLASS, mxREAL));
+    }
+    if (number_of_dims == 3) Output = (float*)mxGetPr(plhs[0] = mxCreateNumericArray(3, dim_array, mxSINGLE_CLASS, mxREAL));
+    
+    NonlDiff_GPU_main(Input, Output, lambda, sigma, iter_numb, tau, penaltytype, dimX, dimY, dimZ);
+}
\ No newline at end of file
diff --git a/Wrappers/Matlab/mex_compile/regularisers_GPU/ROF_TV_GPU.cpp b/Wrappers/Matlab/mex_compile/regularisers_GPU/ROF_TV_GPU.cpp
index 7bbe3af..f60ba7b 100644
--- a/Wrappers/Matlab/mex_compile/regularisers_GPU/ROF_TV_GPU.cpp
+++ b/Wrappers/Matlab/mex_compile/regularisers_GPU/ROF_TV_GPU.cpp
@@ -37,7 +37,6 @@
  *
  * D. Kazantsev, 2016-18
  */
-
 void mexFunction(
         int nlhs, mxArray *plhs[],
         int nrhs, const mxArray *prhs[])
diff --git a/Wrappers/Python/ccpi/filters/regularisers.py b/Wrappers/Python/ccpi/filters/regularisers.py
index e6814e8..eec8c4d 100644
--- a/Wrappers/Python/ccpi/filters/regularisers.py
+++ b/Wrappers/Python/ccpi/filters/regularisers.py
@@ -2,8 +2,8 @@
 script which assigns a proper device core function based on a flag ('cpu' or 'gpu')
 """
 
-from ccpi.filters.cpu_regularisers_cython import TV_ROF_CPU, TV_FGP_CPU, TV_SB_CPU, dTV_FGP_CPU, TNV_CPU 
-from ccpi.filters.gpu_regularisers import TV_ROF_GPU, TV_FGP_GPU, TV_SB_GPU, dTV_FGP_GPU
+from ccpi.filters.cpu_regularisers_cython import TV_ROF_CPU, TV_FGP_CPU, TV_SB_CPU, dTV_FGP_CPU, TNV_CPU, NDF_CPU
+from ccpi.filters.gpu_regularisers import TV_ROF_GPU, TV_FGP_GPU, TV_SB_GPU, dTV_FGP_GPU, NDF_GPU
 
 def ROF_TV(inputData, regularisation_parameter, iterations,
                      time_marching_parameter,device='cpu'):
@@ -91,3 +91,22 @@ def TNV(inputData, regularisation_parameter, iterations, tolerance_param):
                      regularisation_parameter,
                      iterations, 
                      tolerance_param)
+def NDF(inputData, regularisation_parameter, edge_parameter, iterations,
+                     time_marching_parameter, penalty_type, device='cpu'):
+    if device == 'cpu':
+        return NDF_CPU(inputData,
+                     regularisation_parameter,
+                     edge_parameter,
+                     iterations, 
+                     time_marching_parameter,
+                     penalty_type)
+    elif device == 'gpu':
+        return NDF_GPU(inputData,
+                     regularisation_parameter,
+                     edge_parameter,
+                     iterations, 
+                     time_marching_parameter,
+                     penalty_type)
+    else:
+        raise ValueError('Unknown device {0}. Expecting gpu or cpu'\
+                         .format(device))
diff --git a/Wrappers/Python/demos/demo_cpu_regularisers.py b/Wrappers/Python/demos/demo_cpu_regularisers.py
index 7443b83..3567f91 100644
--- a/Wrappers/Python/demos/demo_cpu_regularisers.py
+++ b/Wrappers/Python/demos/demo_cpu_regularisers.py
@@ -12,7 +12,7 @@ import matplotlib.pyplot as plt
 import numpy as np
 import os
 import timeit
-from ccpi.filters.regularisers import ROF_TV, FGP_TV, SB_TV, FGP_dTV, TNV
+from ccpi.filters.regularisers import ROF_TV, FGP_TV, SB_TV, FGP_dTV, TNV, NDF
 from qualitymetrics import rmse
 ###############################################################################
 def printParametersToString(pars):
@@ -190,11 +190,58 @@ plt.title('{}'.format('CPU results'))
 
 
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
-print ("_____________FGP-dTV (2D)__________________")
+print ("________________NDF (2D)___________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
 fig = plt.figure(4)
+plt.suptitle('Performance of NDF regulariser using the CPU')
+a=fig.add_subplot(1,2,1)
+a.set_title('Noisy Image')
+imgplot = plt.imshow(u0,cmap="gray")
+
+# set parameters
+pars = {'algorithm' : NDF, \
+        'input' : u0,\
+        'regularisation_parameter':0.06, \
+        'edge_parameter':0.04,\
+        'number_of_iterations' :1000 ,\
+        'time_marching_parameter':0.025,\
+        'penalty_type':1
+        }
+        
+print ("#############NDF CPU################")
+start_time = timeit.default_timer()
+ndf_cpu = NDF(pars['input'], 
+              pars['regularisation_parameter'],
+              pars['edge_parameter'], 
+              pars['number_of_iterations'],
+              pars['time_marching_parameter'], 
+              pars['penalty_type'],'cpu')  
+             
+rms = rmse(Im, ndf_cpu)
+pars['rmse'] = rms
+
+txtstr = printParametersToString(pars)
+txtstr += "%s = %.3fs" % ('elapsed time',timeit.default_timer() - start_time)
+print (txtstr)
+a=fig.add_subplot(1,2,2)
+
+# these are matplotlib.patch.Patch properties
+props = dict(boxstyle='round', facecolor='wheat', alpha=0.75)
+# place a text box in upper left in axes coords
+a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
+         verticalalignment='top', bbox=props)
+imgplot = plt.imshow(ndf_cpu, cmap="gray")
+plt.title('{}'.format('CPU results'))
+
+
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+print ("_____________FGP-dTV (2D)__________________")
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+
+## plot 
+fig = plt.figure(5)
 plt.suptitle('Performance of FGP-dTV regulariser using the CPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
@@ -247,7 +294,7 @@ print ("__________Total nuclear Variation__________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(5)
+fig = plt.figure(6)
 plt.suptitle('Performance of TNV regulariser using the CPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
@@ -321,7 +368,7 @@ print ("_______________ROF-TV (3D)_________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(6)
+fig = plt.figure(7)
 plt.suptitle('Performance of ROF-TV regulariser using the CPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy 15th slice of a volume')
@@ -361,7 +408,7 @@ print ("_______________FGP-TV (3D)__________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(7)
+fig = plt.figure(8)
 plt.suptitle('Performance of FGP-TV regulariser using the CPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
@@ -410,7 +457,7 @@ print ("_______________SB-TV (3D)_________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(8)
+fig = plt.figure(9)
 plt.suptitle('Performance of SB-TV regulariser using the CPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
@@ -451,13 +498,58 @@ a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
 imgplot = plt.imshow(sb_cpu3D[10,:,:], cmap="gray")
 plt.title('{}'.format('Recovered volume on the CPU using SB-TV'))
 
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+print ("________________NDF (3D)___________________")
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+
+## plot 
+fig = plt.figure(10)
+plt.suptitle('Performance of NDF regulariser using the CPU')
+a=fig.add_subplot(1,2,1)
+a.set_title('Noisy volume')
+imgplot = plt.imshow(noisyVol[10,:,:],cmap="gray")
+
+# set parameters
+pars = {'algorithm' : NDF, \
+        'input' : noisyVol,\
+        'regularisation_parameter':0.06, \
+        'edge_parameter':0.04,\
+        'number_of_iterations' :1000 ,\
+        'time_marching_parameter':0.025,\
+        'penalty_type':  1
+        }
+        
+print ("#############NDF CPU################")
+start_time = timeit.default_timer()
+ndf_cpu3D = NDF(pars['input'], 
+              pars['regularisation_parameter'],
+              pars['edge_parameter'], 
+              pars['number_of_iterations'],
+              pars['time_marching_parameter'], 
+              pars['penalty_type'])  
+             
+rms = rmse(idealVol, ndf_cpu3D)
+pars['rmse'] = rms
+
+txtstr = printParametersToString(pars)
+txtstr += "%s = %.3fs" % ('elapsed time',timeit.default_timer() - start_time)
+print (txtstr)
+a=fig.add_subplot(1,2,2)
+
+# these are matplotlib.patch.Patch properties
+props = dict(boxstyle='round', facecolor='wheat', alpha=0.75)
+# place a text box in upper left in axes coords
+a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
+         verticalalignment='top', bbox=props)
+imgplot = plt.imshow(ndf_cpu3D[10,:,:], cmap="gray")
+plt.title('{}'.format('Recovered volume on the CPU using NDF iterations'))
 
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 print ("_______________FGP-dTV (3D)__________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(9)
+fig = plt.figure(11)
 plt.suptitle('Performance of FGP-dTV regulariser using the CPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
diff --git a/Wrappers/Python/demos/demo_cpu_vs_gpu_regularisers.py b/Wrappers/Python/demos/demo_cpu_vs_gpu_regularisers.py
index d8e2da7..05db23e 100644
--- a/Wrappers/Python/demos/demo_cpu_vs_gpu_regularisers.py
+++ b/Wrappers/Python/demos/demo_cpu_vs_gpu_regularisers.py
@@ -12,7 +12,7 @@ import matplotlib.pyplot as plt
 import numpy as np
 import os
 import timeit
-from ccpi.filters.regularisers import ROF_TV, FGP_TV, SB_TV, FGP_dTV
+from ccpi.filters.regularisers import ROF_TV, FGP_TV, SB_TV, FGP_dTV, NDF
 from qualitymetrics import rmse
 ###############################################################################
 def printParametersToString(pars):
@@ -306,11 +306,98 @@ else:
 
 
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
-print ("____________FGP-dTV bench___________________")
+print ("_______________NDF bench___________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
 fig = plt.figure(4)
+plt.suptitle('Comparison of NDF regulariser using CPU and GPU implementations')
+a=fig.add_subplot(1,4,1)
+a.set_title('Noisy Image')
+imgplot = plt.imshow(u0,cmap="gray")
+
+# set parameters
+pars = {'algorithm' : NDF, \
+        'input' : u0,\
+        'regularisation_parameter':0.06, \
+        'edge_parameter':0.04,\
+        'number_of_iterations' :1000 ,\
+        'time_marching_parameter':0.025,\
+        'penalty_type':  1
+        }
+        
+print ("#############NDF CPU####################")
+start_time = timeit.default_timer()
+ndf_cpu = NDF(pars['input'], 
+              pars['regularisation_parameter'],
+              pars['edge_parameter'], 
+              pars['number_of_iterations'],
+              pars['time_marching_parameter'], 
+              pars['penalty_type'],'cpu')
+             
+rms = rmse(Im, ndf_cpu)
+pars['rmse'] = rms
+
+txtstr = printParametersToString(pars)
+txtstr += "%s = %.3fs" % ('elapsed time',timeit.default_timer() - start_time)
+print (txtstr)
+a=fig.add_subplot(1,4,2)
+
+# these are matplotlib.patch.Patch properties
+props = dict(boxstyle='round', facecolor='wheat', alpha=0.75)
+# place a text box in upper left in axes coords
+a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
+         verticalalignment='top', bbox=props)
+imgplot = plt.imshow(ndf_cpu, cmap="gray")
+plt.title('{}'.format('CPU results'))
+
+
+print ("##############NDF GPU##################")
+start_time = timeit.default_timer()
+ndf_gpu = NDF(pars['input'], 
+              pars['regularisation_parameter'],
+              pars['edge_parameter'], 
+              pars['number_of_iterations'],
+              pars['time_marching_parameter'], 
+              pars['penalty_type'],'gpu')
+             
+rms = rmse(Im, ndf_gpu)
+pars['rmse'] = rms
+pars['algorithm'] = NDF
+txtstr = printParametersToString(pars)
+txtstr += "%s = %.3fs" % ('elapsed time',timeit.default_timer() - start_time)
+print (txtstr)
+a=fig.add_subplot(1,4,3)
+
+# these are matplotlib.patch.Patch properties
+props = dict(boxstyle='round', facecolor='wheat', alpha=0.75)
+# place a text box in upper left in axes coords
+a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
+         verticalalignment='top', bbox=props)
+imgplot = plt.imshow(ndf_gpu, cmap="gray")
+plt.title('{}'.format('GPU results'))
+
+print ("--------Compare the results--------")
+tolerance = 1e-05
+diff_im = np.zeros(np.shape(rof_cpu))
+diff_im = abs(ndf_cpu - ndf_gpu)
+diff_im[diff_im > tolerance] = 1
+a=fig.add_subplot(1,4,4)
+imgplot = plt.imshow(diff_im, vmin=0, vmax=1, cmap="gray")
+plt.title('{}'.format('Pixels larger threshold difference'))
+if (diff_im.sum() > 1):
+    print ("Arrays do not match!")
+else:
+    print ("Arrays match")
+
+
+
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+print ("____________FGP-dTV bench___________________")
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+
+## plot 
+fig = plt.figure(5)
 plt.suptitle('Comparison of FGP-dTV regulariser using CPU and GPU implementations')
 a=fig.add_subplot(1,4,1)
 a.set_title('Noisy Image')
diff --git a/Wrappers/Python/demos/demo_gpu_regularisers.py b/Wrappers/Python/demos/demo_gpu_regularisers.py
index 25d8d85..b873700 100644
--- a/Wrappers/Python/demos/demo_gpu_regularisers.py
+++ b/Wrappers/Python/demos/demo_gpu_regularisers.py
@@ -12,7 +12,7 @@ import matplotlib.pyplot as plt
 import numpy as np
 import os
 import timeit
-from ccpi.filters.regularisers import ROF_TV, FGP_TV, SB_TV, FGP_dTV
+from ccpi.filters.regularisers import ROF_TV, FGP_TV, SB_TV, FGP_dTV, NDF
 from qualitymetrics import rmse
 ###############################################################################
 def printParametersToString(pars):
@@ -50,7 +50,7 @@ u0 = u0.astype('float32')
 u_ref = u_ref.astype('float32')
 
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
-print ("____________ROF-TV bench___________________")
+print ("____________ROF-TV regulariser_____________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
@@ -92,7 +92,7 @@ plt.title('{}'.format('GPU results'))
 
 
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
-print ("____________FGP-TV bench___________________")
+print ("____________FGP-TV regulariser_____________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
@@ -141,7 +141,7 @@ plt.title('{}'.format('GPU results'))
 
 
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
-print ("____________SB-TV bench___________________")
+print ("____________SB-TV regulariser______________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
@@ -186,12 +186,60 @@ a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
 imgplot = plt.imshow(sb_gpu, cmap="gray")
 plt.title('{}'.format('GPU results'))
 
+
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
-print ("____________FGP-dTV bench___________________")
+print ("_______________NDF regulariser_____________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
 fig = plt.figure(4)
+plt.suptitle('Performance of the NDF regulariser using the GPU')
+a=fig.add_subplot(1,2,1)
+a.set_title('Noisy Image')
+imgplot = plt.imshow(u0,cmap="gray")
+
+# set parameters
+pars = {'algorithm' : NDF, \
+        'input' : u0,\
+        'regularisation_parameter':0.06, \
+        'edge_parameter':0.04,\
+        'number_of_iterations' :1000 ,\
+        'time_marching_parameter':0.025,\
+        'penalty_type':  1
+        }
+
+print ("##############NDF GPU##################")
+start_time = timeit.default_timer()
+ndf_gpu = NDF(pars['input'], 
+              pars['regularisation_parameter'],
+              pars['edge_parameter'], 
+              pars['number_of_iterations'],
+              pars['time_marching_parameter'], 
+              pars['penalty_type'],'gpu')  
+             
+rms = rmse(Im, ndf_gpu)
+pars['rmse'] = rms
+pars['algorithm'] = NDF
+txtstr = printParametersToString(pars)
+txtstr += "%s = %.3fs" % ('elapsed time',timeit.default_timer() - start_time)
+print (txtstr)
+a=fig.add_subplot(1,2,2)
+
+# these are matplotlib.patch.Patch properties
+props = dict(boxstyle='round', facecolor='wheat', alpha=0.75)
+# place a text box in upper left in axes coords
+a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
+         verticalalignment='top', bbox=props)
+imgplot = plt.imshow(ndf_gpu, cmap="gray")
+plt.title('{}'.format('GPU results'))
+
+
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+print ("____________FGP-dTV bench___________________")
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+
+## plot 
+fig = plt.figure(5)
 plt.suptitle('Performance of the FGP-dTV regulariser using the GPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
@@ -266,7 +314,7 @@ print ("_______________ROF-TV (3D)_________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(5)
+fig = plt.figure(6)
 plt.suptitle('Performance of ROF-TV regulariser using the GPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy 15th slice of a volume')
@@ -306,7 +354,7 @@ print ("_______________FGP-TV (3D)__________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(6)
+fig = plt.figure(7)
 plt.suptitle('Performance of FGP-TV regulariser using the GPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
@@ -354,7 +402,7 @@ print ("_______________SB-TV (3D)__________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(7)
+fig = plt.figure(8)
 plt.suptitle('Performance of SB-TV regulariser using the GPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
@@ -395,12 +443,60 @@ a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
 imgplot = plt.imshow(sb_gpu3D[10,:,:], cmap="gray")
 plt.title('{}'.format('Recovered volume on the GPU using SB-TV'))
 
+
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+print ("_______________NDF-TV (3D)_________________")
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+
+## plot 
+fig = plt.figure(9)
+plt.suptitle('Performance of NDF regulariser using the GPU')
+a=fig.add_subplot(1,2,1)
+a.set_title('Noisy Image')
+imgplot = plt.imshow(noisyVol[10,:,:],cmap="gray")
+
+# set parameters
+pars = {'algorithm' : NDF, \
+        'input' : noisyVol,\
+        'regularisation_parameter':0.06, \
+        'edge_parameter':0.04,\
+        'number_of_iterations' :1000 ,\
+        'time_marching_parameter':0.025,\
+        'penalty_type':  1
+        }
+
+print ("#############NDF GPU####################")
+start_time = timeit.default_timer()
+ndf_gpu3D = NDF(pars['input'], 
+              pars['regularisation_parameter'],
+              pars['edge_parameter'], 
+              pars['number_of_iterations'],
+              pars['time_marching_parameter'], 
+              pars['penalty_type'],'gpu')
+
+rms = rmse(idealVol, ndf_gpu3D)
+pars['rmse'] = rms
+
+txtstr = printParametersToString(pars)
+txtstr += "%s = %.3fs" % ('elapsed time',timeit.default_timer() - start_time)
+print (txtstr)
+a=fig.add_subplot(1,2,2)
+
+# these are matplotlib.patch.Patch properties
+props = dict(boxstyle='round', facecolor='wheat', alpha=0.75)
+# place a text box in upper left in axes coords
+a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
+         verticalalignment='top', bbox=props)
+imgplot = plt.imshow(ndf_gpu3D[10,:,:], cmap="gray")
+plt.title('{}'.format('Recovered volume on the GPU using NDF'))
+
+
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 print ("_______________FGP-dTV (3D)________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 
 ## plot 
-fig = plt.figure(8)
+fig = plt.figure(10)
 plt.suptitle('Performance of FGP-dTV regulariser using the GPU')
 a=fig.add_subplot(1,2,1)
 a.set_title('Noisy Image')
diff --git a/Wrappers/Python/setup-regularisers.py.in b/Wrappers/Python/setup-regularisers.py.in
index 0681cc4..b900efe 100644
--- a/Wrappers/Python/setup-regularisers.py.in
+++ b/Wrappers/Python/setup-regularisers.py.in
@@ -37,6 +37,7 @@ extra_include_dirs += [os.path.join(".." , ".." , "Core"),
                        os.path.join(".." , ".." , "Core",  "regularisers_GPU" , "TV_FGP" ) , 
                        os.path.join(".." , ".." , "Core",  "regularisers_GPU" , "TV_ROF" ) , 
                        os.path.join(".." , ".." , "Core",  "regularisers_GPU" , "TV_SB" ) ,
+                       os.path.join(".." , ".." , "Core",  "regularisers_GPU" , "NDF" ) ,
                        os.path.join(".." , ".." , "Core",  "regularisers_GPU" , "dTV_FGP" ) , 
 						   "."]
 
diff --git a/Wrappers/Python/src/cpu_regularisers.pyx b/Wrappers/Python/src/cpu_regularisers.pyx
index abbf3b0..7ed8fa1 100644
--- a/Wrappers/Python/src/cpu_regularisers.pyx
+++ b/Wrappers/Python/src/cpu_regularisers.pyx
@@ -21,10 +21,10 @@ cimport numpy as np
 cdef extern float TV_ROF_CPU_main(float *Input, float *Output, float lambdaPar, int iterationsNumb, float tau, int dimX, int dimY, int dimZ);
 cdef extern float TV_FGP_CPU_main(float *Input, float *Output, float lambdaPar, int iterationsNumb, float epsil, int methodTV, int nonneg, int printM, int dimX, int dimY, int dimZ);
 cdef extern float SB_TV_CPU_main(float *Input, float *Output, float lambdaPar, int iterationsNumb, float epsil, int methodTV, int printM, int dimX, int dimY, int dimZ);
+cdef extern float Diffusion_CPU_main(float *Input, float *Output, float lambdaPar, float sigmaPar, int iterationsNumb, float tau, int penaltytype, int dimX, int dimY, int dimZ);
 cdef extern float TNV_CPU_main(float *Input, float *u, float lambdaPar, int maxIter, float tol, int dimX, int dimY, int dimZ);
 cdef extern float dTV_FGP_CPU_main(float *Input, float *InputRef, float *Output, float lambdaPar, int iterationsNumb, float epsil, float eta, int methodTV, int nonneg, int printM, int dimX, int dimY, int dimZ);
 
-
 #****************************************************************#
 #********************** Total-variation ROF *********************#
 #****************************************************************#
@@ -275,3 +275,47 @@ def TNV_3D(np.ndarray[np.float32_t, ndim=3, mode="c"] inputData,
     # Run TNV iterations for 3D (X,Y,Channels) data 
     TNV_CPU_main(&inputData[0,0,0], &outputData[0,0,0], regularisation_parameter, iterationsNumb, tolerance_param, dims[2], dims[1], dims[0])
     return outputData
+#****************************************************************#
+#***************Nonlinear (Isotropic) Diffusion******************#
+#****************************************************************#
+def NDF_CPU(inputData, regularisation_parameter, edge_parameter, iterationsNumb,time_marching_parameter, penalty_type):
+    if inputData.ndim == 2:
+        return NDF_2D(inputData, regularisation_parameter, edge_parameter, iterationsNumb, time_marching_parameter, penalty_type)
+    elif inputData.ndim == 3:
+        return NDF_3D(inputData, regularisation_parameter, edge_parameter, iterationsNumb, time_marching_parameter, penalty_type)
+
+def NDF_2D(np.ndarray[np.float32_t, ndim=2, mode="c"] inputData, 
+                     float regularisation_parameter,
+                     float edge_parameter,
+                     int iterationsNumb,                     
+                     float time_marching_parameter,
+                     int penalty_type):
+    cdef long dims[2]
+    dims[0] = inputData.shape[0]
+    dims[1] = inputData.shape[1]
+    
+    cdef np.ndarray[np.float32_t, ndim=2, mode="c"] outputData = \
+            np.zeros([dims[0],dims[1]], dtype='float32')   
+    
+    # Run Nonlinear Diffusion iterations for 2D data 
+    Diffusion_CPU_main(&inputData[0,0], &outputData[0,0], regularisation_parameter, edge_parameter, iterationsNumb, time_marching_parameter, penalty_type, dims[0], dims[1], 1)    
+    return outputData
+            
+def NDF_3D(np.ndarray[np.float32_t, ndim=3, mode="c"] inputData, 
+                     float regularisation_parameter,
+                     float edge_parameter,
+                     int iterationsNumb,                     
+                     float time_marching_parameter,
+                     int penalty_type):
+    cdef long dims[3]
+    dims[0] = inputData.shape[0]
+    dims[1] = inputData.shape[1]
+    dims[2] = inputData.shape[2]
+    
+    cdef np.ndarray[np.float32_t, ndim=3, mode="c"] outputData = \
+            np.zeros([dims[0],dims[1],dims[2]], dtype='float32')
+    
+    # Run Nonlinear Diffusion iterations for  3D data 
+    Diffusion_CPU_main(&inputData[0,0,0], &outputData[0,0,0], regularisation_parameter, edge_parameter, iterationsNumb, time_marching_parameter, penalty_type, dims[2], dims[1], dims[0])
+
+    return outputData
diff --git a/Wrappers/Python/src/gpu_regularisers.pyx b/Wrappers/Python/src/gpu_regularisers.pyx
index 36eec95..b0775054 100644
--- a/Wrappers/Python/src/gpu_regularisers.pyx
+++ b/Wrappers/Python/src/gpu_regularisers.pyx
@@ -21,6 +21,7 @@ cimport numpy as np
 cdef extern void TV_ROF_GPU_main(float* Input, float* Output, float lambdaPar, int iter, float tau, int N, int M, int Z);
 cdef extern void TV_FGP_GPU_main(float *Input, float *Output, float lambdaPar, int iter, float epsil, int methodTV, int nonneg, int printM, int N, int M, int Z);
 cdef extern void TV_SB_GPU_main(float *Input, float *Output, float lambdaPar, int iter, float epsil, int methodTV, int printM, int N, int M, int Z);
+cdef extern void NonlDiff_GPU_main(float *Input, float *Output, float lambdaPar, float sigmaPar, int iterationsNumb, float tau, int penaltytype, int N, int M, int Z);
 cdef extern void dTV_FGP_GPU_main(float *Input, float *InputRef, float *Output, float lambdaPar, int iterationsNumb, float epsil, float eta, int methodTV, int nonneg, int printM, int N, int M, int Z);
 
 # Total-variation Rudin-Osher-Fatemi (ROF)
@@ -114,6 +115,27 @@ def dTV_FGP_GPU(inputData,
                      methodTV,
                      nonneg,
                      printM)
+# Nonlocal Isotropic Diffusion (NDF)
+def NDF_GPU(inputData,
+                     regularisation_parameter,
+                     edge_parameter,
+                     iterations, 
+                     time_marching_parameter,
+                     penalty_type):
+    if inputData.ndim == 2:
+        return NDF_GPU_2D(inputData,
+                     regularisation_parameter,
+                     edge_parameter,
+                     iterations, 
+                     time_marching_parameter,
+                     penalty_type)
+    elif inputData.ndim == 3:
+        return NDF_GPU_3D(inputData,
+                     regularisation_parameter,
+                     edge_parameter,
+                     iterations, 
+                     time_marching_parameter,
+                     penalty_type)                     
 #****************************************************************#
 #********************** Total-variation ROF *********************#
 #****************************************************************#
@@ -336,3 +358,48 @@ def FGPdTV3D(np.ndarray[np.float32_t, ndim=3, mode="c"] inputData,
                        printM,
                        dims[2], dims[1], dims[0]);
     return outputData 
+
+#****************************************************************#
+#***************Nonlinear (Isotropic) Diffusion******************#
+#****************************************************************#
+def NDF_GPU_2D(np.ndarray[np.float32_t, ndim=2, mode="c"] inputData, 
+                     float regularisation_parameter,
+                     float edge_parameter,
+                     int iterationsNumb,                     
+                     float time_marching_parameter,
+                     int penalty_type):
+    cdef long dims[2]
+    dims[0] = inputData.shape[0]
+    dims[1] = inputData.shape[1]
+    
+    cdef np.ndarray[np.float32_t, ndim=2, mode="c"] outputData = \
+            np.zeros([dims[0],dims[1]], dtype='float32')
+    
+    #rangecheck = penalty_type < 1 and penalty_type > 3
+    #if not rangecheck:
+#        raise ValueError('Choose penalty type as 1 for Huber, 2 - Perona-Malik, 3 - Tukey Biweight')
+    
+    # Run Nonlinear Diffusion iterations for 2D data 
+    # Running CUDA code here  
+    NonlDiff_GPU_main(&inputData[0,0], &outputData[0,0], regularisation_parameter, edge_parameter, iterationsNumb, time_marching_parameter, penalty_type, dims[0], dims[1], 1)    
+    return outputData
+            
+def NDF_GPU_3D(np.ndarray[np.float32_t, ndim=3, mode="c"] inputData, 
+                     float regularisation_parameter,
+                     float edge_parameter,
+                     int iterationsNumb,                     
+                     float time_marching_parameter,
+                     int penalty_type):
+    cdef long dims[3]
+    dims[0] = inputData.shape[0]
+    dims[1] = inputData.shape[1]
+    dims[2] = inputData.shape[2]
+    
+    cdef np.ndarray[np.float32_t, ndim=3, mode="c"] outputData = \
+            np.zeros([dims[0],dims[1],dims[2]], dtype='float32')    
+       
+    # Run Nonlinear Diffusion iterations for  3D data 
+    # Running CUDA code here  
+    NonlDiff_GPU_main(&inputData[0,0,0], &outputData[0,0,0], regularisation_parameter, edge_parameter, iterationsNumb, time_marching_parameter, penalty_type, dims[2], dims[1], dims[0])
+
+    return outputData