cythonised nltv and updated demo, readme, bash run added

3 files changed, 145 insertions, 19 deletions

diff --git a/Wrappers/Python/ccpi/filters/regularisers.py b/Wrappers/Python/ccpi/filters/regularisers.py
index c3c3c7e..bf7e23c 100644
--- a/Wrappers/Python/ccpi/filters/regularisers.py
+++ b/Wrappers/Python/ccpi/filters/regularisers.py
@@ -2,7 +2,7 @@
 script which assigns a proper device core function based on a flag ('cpu' or 'gpu')
 """
 
-from ccpi.filters.cpu_regularisers import TV_ROF_CPU, TV_FGP_CPU, TV_SB_CPU, dTV_FGP_CPU, TNV_CPU, NDF_CPU, Diff4th_CPU, TGV_CPU, LLT_ROF_CPU, PATCHSEL_CPU
+from ccpi.filters.cpu_regularisers import TV_ROF_CPU, TV_FGP_CPU, TV_SB_CPU, dTV_FGP_CPU, TNV_CPU, NDF_CPU, Diff4th_CPU, TGV_CPU, LLT_ROF_CPU, PATCHSEL_CPU, NLTV_CPU
 try:
     from ccpi.filters.gpu_regularisers import TV_ROF_GPU, TV_FGP_GPU, TV_SB_GPU, dTV_FGP_GPU, NDF_GPU, Diff4th_GPU, TGV_GPU, LLT_ROF_GPU
     gpu_enabled = True
@@ -145,7 +145,7 @@ def DIFF4th(inputData, regularisation_parameter, edge_parameter, iterations,
         raise ValueError('Unknown device {0}. Expecting gpu or cpu'\
                          .format(device))
         
-def PatchSelect_CPU(inputData, searchwindow, patchwindow, neighbours, edge_parameter, device='cpu'):
+def PatchSelect(inputData, searchwindow, patchwindow, neighbours, edge_parameter, device='cpu'):
     if device == 'cpu':
         return PATCHSEL_CPU(inputData,
                      searchwindow,
@@ -159,7 +159,16 @@ def PatchSelect_CPU(inputData, searchwindow, patchwindow, neighbours, edge_param
     	    raise ValueError ('GPU is not available')
         raise ValueError('Unknown device {0}. Expecting gpu or cpu'\
                          .format(device))
-        
+
+def NLTV(inputData, H_i, H_j, H_k, Weights, regularisation_parameter, iterations):
+    return NLTV_CPU(inputData,
+                     H_i,
+                     H_j,
+                     H_k, 
+                     Weights,
+                     regularisation_parameter,
+                     iterations)
+
 def TGV(inputData, regularisation_parameter, alpha1, alpha0, iterations,
                      LipshitzConst, device='cpu'):
     if device == 'cpu':
diff --git a/Wrappers/Python/demos/demo_cpu_regularisers.py b/Wrappers/Python/demos/demo_cpu_regularisers.py
index e99b271..31e4cad 100644
--- a/Wrappers/Python/demos/demo_cpu_regularisers.py
+++ b/Wrappers/Python/demos/demo_cpu_regularisers.py
@@ -13,6 +13,7 @@ import numpy as np
 import os
 import timeit
 from ccpi.filters.regularisers import ROF_TV, FGP_TV, SB_TV, TGV, LLT_ROF, FGP_dTV, TNV, NDF, DIFF4th
+from ccpi.filters.regularisers import PatchSelect, NLTV
 from qualitymetrics import rmse
 ###############################################################################
 def printParametersToString(pars):
@@ -350,7 +351,7 @@ a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
 imgplot = plt.imshow(ndf_cpu, cmap="gray")
 plt.title('{}'.format('CPU results'))
 
-
+#%%
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 print ("___Anisotropic Diffusion 4th Order (2D)____")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
@@ -395,7 +396,71 @@ a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
 imgplot = plt.imshow(diff4_cpu, cmap="gray")
 plt.title('{}'.format('CPU results'))
 
+#%%
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+print ("___Nonlocal patches pre-calculation____")
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+# set parameters
+pars = {'algorithm' : PatchSelect, \
+        'input' : u0,\
+        'searchwindow': 7, \
+        'patchwindow': 2,\
+        'neighbours' : 15 ,\
+        'edge_parameter':0.23}
+
+H_i, H_j, Weights = PatchSelect(pars['input'], 
+              pars['searchwindow'],
+              pars['patchwindow'], 
+              pars['neighbours'],
+              pars['edge_parameter'],'cpu')
+
+#%%
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+print ("___Nonlocal Total Variation penalty____")
+print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
+## plot 
+fig = plt.figure()
+plt.suptitle('Performance of NLTV regulariser using the CPU')
+a=fig.add_subplot(1,2,1)
+a.set_title('Noisy Image')
+imgplot = plt.imshow(u0,cmap="gray")
+
+pars2 = {'algorithm' : NLTV, \
+        'input' : u0,\
+        'H_i': H_i, \
+        'H_j': H_j,\
+        'H_k' : 0,\
+        'Weights' : Weights,\
+        'regularisation_parameter': 0.085,\
+        'iterations': 2
+        }
+#%%
+start_time = timeit.default_timer()
+nltv_cpu = NLTV(pars2['input'], 
+              pars2['H_i'],
+              pars2['H_j'], 
+              pars2['H_k'],
+              pars2['Weights'],
+              pars2['regularisation_parameter'],
+              pars2['iterations'])
+
+rms = rmse(Im, nltv_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(nltv_cpu, cmap="gray")
+plt.title('{}'.format('CPU results'))
+
+#%%
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 print ("_____________FGP-dTV (2D)__________________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
@@ -447,7 +512,7 @@ a.text(0.15, 0.25, txtstr, transform=a.transAxes, fontsize=14,
          verticalalignment='top', bbox=props)
 imgplot = plt.imshow(fgp_dtv_cpu, cmap="gray")
 plt.title('{}'.format('CPU results'))
-
+#%%
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
 print ("__________Total nuclear Variation__________")
 print ("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
diff --git a/Wrappers/Python/src/cpu_regularisers.pyx b/Wrappers/Python/src/cpu_regularisers.pyx
index b056bba..e51e6d8 100644
--- a/Wrappers/Python/src/cpu_regularisers.pyx
+++ b/Wrappers/Python/src/cpu_regularisers.pyx
@@ -27,10 +27,9 @@ cdef extern float Diffusion_CPU_main(float *Input, float *Output, float lambdaPa
 cdef extern float Diffus4th_CPU_main(float *Input, float *Output, float lambdaPar, float sigmaPar, int iterationsNumb, float tau, 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);
-cdef extern float PatchSelect_CPU_main(float *Input, unsigned short *H_i, unsigned short *H_j, unsigned short *H_k, float *Weights, long dimX, long dimY, long dimZ, int SearchWindow, int SimilarWin, int NumNeighb, float h);
+cdef extern float PatchSelect_CPU_main(float *Input, unsigned short *H_i, unsigned short *H_j, unsigned short *H_k, float *Weights, int dimX, int dimY, int dimZ, int SearchWindow, int SimilarWin, int NumNeighb, float h, int switchM);
 cdef extern float Nonlocal_TV_CPU_main(float *A_orig, float *Output, unsigned short *H_i, unsigned short *H_j, unsigned short *H_k, float *Weights, int dimX, int dimY, int dimZ, int NumNeighb, float lambdaReg, int IterNumb);
 
-
 cdef extern float Diffusion_Inpaint_CPU_main(float *Input, unsigned char *Mask, float *Output, float lambdaPar, float sigmaPar, int iterationsNumb, float tau, int penaltytype, int dimX, int dimY, int dimZ);
 cdef extern float NonlocalMarching_Inpaint_main(float *Input, unsigned char *M, float *Output, unsigned char *M_upd, int SW_increment, int iterationsNumb, int trigger, int dimX, int dimY, int dimZ);
 cdef extern float TV_energy2D(float *U, float *U0, float *E_val, float lambdaPar, int type, int dimX, int dimY);
@@ -450,7 +449,6 @@ def Diff4th_3D(np.ndarray[np.float32_t, ndim=3, mode="c"] inputData,
 
     return outputData
 
-
 #****************************************************************#
 #***************Patch-based weights calculation******************#
 #****************************************************************#
@@ -458,31 +456,85 @@ def PATCHSEL_CPU(inputData, searchwindow, patchwindow, neighbours, edge_paramete
     if inputData.ndim == 2:
         return PatchSel_2D(inputData, searchwindow, patchwindow, neighbours, edge_parameter)
     elif inputData.ndim == 3:
-        return 1
-# PatchSel_3D(inputData, searchwindow, patchwindow, neighbours, edge_parameter)
-def PatchSel_2D(np.ndarray[np.float32_t, ndim=2, mode="c"] inputData, 
+        return PatchSel_3D(inputData, searchwindow, patchwindow, neighbours, edge_parameter)
+def PatchSel_2D(np.ndarray[np.float32_t, ndim=2, mode="c"] inputData,
                      int searchwindow,
                      int patchwindow,
                      int neighbours,
                      float edge_parameter):
     cdef long dims[3]
-    dims[0] = inputData.shape[0]
-    dims[1] = inputData.shape[1]
-    dims[2] = neighbours
+    dims[0] = neighbours
+    dims[1] = inputData.shape[0]
+    dims[2] = inputData.shape[1]
+    
     
     cdef np.ndarray[np.float32_t, ndim=3, mode="c"] Weights = \
-            np.zeros([dims[0],dims[1],dims[2]], dtype='float32')
+            np.zeros([dims[0], dims[1],dims[2]], dtype='float32')
     
     cdef np.ndarray[np.uint16_t, ndim=3, mode="c"] H_i = \
-            np.zeros([dims[0],dims[1],dims[2]], dtype='uint16 ')
+            np.zeros([dims[0], dims[1],dims[2]], dtype='uint16')
             
     cdef np.ndarray[np.uint16_t, ndim=3, mode="c"] H_j = \
-            np.zeros([dims[0],dims[1],dims[2]], dtype='uint16 ')
+            np.zeros([dims[0], dims[1],dims[2]], dtype='uint16')
 
     # Run patch-based weight selection function
-    PatchSelect_CPU_main(&inputData[0,0], &H_i[0,0,0], &H_j[0,0,0], &H_i[0,0,0], &Weights[0,0,0], dims[1], dims[0], 1, searchwindow, patchwindow,  neighbours,  edge_parameter)
+    PatchSelect_CPU_main(&inputData[0,0], &H_j[0,0,0], &H_i[0,0,0], &H_i[0,0,0], &Weights[0,0,0], dims[2], dims[1], 0, searchwindow, patchwindow,  neighbours,  edge_parameter, 1)
     return H_i, H_j, Weights
-          
+
+def PatchSel_3D(np.ndarray[np.float32_t, ndim=3, mode="c"] inputData,
+                     int searchwindow,
+                     int patchwindow,
+                     int neighbours,
+                     float edge_parameter):
+    cdef long dims[4]
+    dims[0] = inputData.shape[0]
+    dims[1] = inputData.shape[1]
+    dims[2] = inputData.shape[2]
+    dims[3] = neighbours
+    
+    cdef np.ndarray[np.float32_t, ndim=4, mode="c"] Weights = \
+            np.zeros([dims[3],dims[0],dims[1],dims[2]], dtype='float32')
+    
+    cdef np.ndarray[np.uint16_t, ndim=4, mode="c"] H_i = \
+            np.zeros([dims[3],dims[0],dims[1],dims[2]], dtype='uint16')
+            
+    cdef np.ndarray[np.uint16_t, ndim=4, mode="c"] H_j = \
+            np.zeros([dims[3],dims[0],dims[1],dims[2]], dtype='uint16')
+            
+    cdef np.ndarray[np.uint16_t, ndim=4, mode="c"] H_k = \
+            np.zeros([dims[3],dims[0],dims[1],dims[2]], dtype='uint16')
+
+    # Run patch-based weight selection function
+    PatchSelect_CPU_main(&inputData[0,0,0], &H_i[0,0,0,0], &H_j[0,0,0,0], &H_k[0,0,0,0], &Weights[0,0,0,0], dims[2], dims[1], dims[0], searchwindow, patchwindow,  neighbours, edge_parameter, 1)
+    return H_i, H_j, H_k, Weights
+
+
+#****************************************************************#
+#***************Non-local Total Variation******************#
+#****************************************************************#
+def NLTV_CPU(inputData, H_i, H_j, H_k, Weights, regularisation_parameter, iterations):
+    if inputData.ndim == 2:
+        return NLTV_2D(inputData, H_i, H_j, Weights, regularisation_parameter, iterations)
+    elif inputData.ndim == 3:
+        return 1
+def NLTV_2D(np.ndarray[np.float32_t, ndim=2, mode="c"] inputData,
+                     np.ndarray[np.uint16_t, ndim=3, mode="c"] H_i,
+                     np.ndarray[np.uint16_t, ndim=3, mode="c"] H_j,
+                     np.ndarray[np.float32_t, ndim=3, mode="c"] Weights,
+                     float regularisation_parameter,
+                     int iterations):
+
+    cdef long dims[2]
+    dims[0] = inputData.shape[0]
+    dims[1] = inputData.shape[1]
+    neighbours = H_i.shape[0]
+    
+    cdef np.ndarray[np.float32_t, ndim=2, mode="c"] outputData = \
+            np.zeros([dims[0],dims[1]], dtype='float32')
+    
+    # Run nonlocal TV regularisation
+    Nonlocal_TV_CPU_main(&inputData[0,0], &outputData[0,0], &H_i[0,0,0], &H_j[0,0,0], &H_i[0,0,0], &Weights[0,0,0], dims[1], dims[0], 0, neighbours, regularisation_parameter, iterations)
+    return outputData
 
 #*********************Inpainting WITH****************************#
 #***************Nonlinear (Isotropic) Diffusion******************#