def Graph_Matcher(
full_graph,
ROI_graph,
):
if full_graph.dtype == 'u2':
full_graph = (full_graph // 256).astype('u1')
aligned_ROI, h = Affine_Core_Point_Equal(ROI_graph,
full_graph,
targ_gain=1)
aligned_ROI = (aligned_ROI // 256).astype('u1')
# Here we get affine matrix h and aligned ROI graph.
merged_graph = cv2.cvtColor(full_graph, cv2.COLOR_GRAY2RGB).astype('f8')
location_graph = cv2.cvtColor(full_graph, cv2.COLOR_GRAY2RGB).astype(
'f8') # Define location here.
merged_graph[:, :, 1] += aligned_ROI
merged_graph = np.clip(merged_graph, 0, 255).astype('u1')
# Then we annotate graph boulder in graph.
rectangle = np.zeros(np.shape(ROI_graph), dtype='u1')
rectangle = Graph_Tools.Boulder_Fill(rectangle, [3, 3, 3, 3], 255)
ROI_boulder = cv2.warpPerspective(rectangle, h, (512, 512))
location_graph[:, :, 2] += ROI_boulder
location_graph = np.clip(location_graph, 0, 255).astype('u1')
return merged_graph, location_graph, ROI_boulder, h
def Cell_Find_From_Graph(
input_graph,
find_thres = 'otsu',
max_pix = 1000,
min_pix = 20,
shape_boulder = [20,20,20,20],
sharp_gauss = ([7,7],1.5),
back_gauss = ([15,15],7),
size_limit = 20
):
"""
Find Cell From Graph,this graph can be average intensity or On-Off sub map.
Parameters
----------
input_graph : (2D Array)
Input graph. Use this graph to find cell.
find_thres : (float or 'otsu'), optional
How many std we use in finding cells. The default is 'otsu', meaning we use ootsu method to find threshold here.
max_pix : (int), optional
Max cell area. Bigger than this will be ignored. The default is 1000.
min_pix : (int), optional
Mininum cell area. Smaller than this will be ignored. The default is 20.
shape_boulder : (4 element list), optional
Cell find boulder, usually same as align range. The default is 20.
[Up,Down,Left,Right]
sharp_gauss : (turple), optional
Sharp gaussian blur. Cell data will be found there. The default is ([7,7],1.5).
blur_gauss : (turple), optional
Coarse gaussian blur, this value determine local average areas. The default is ([15,15],7).
size_limit : (int), optional
Width and Length limitarion. Area longer than this will be ignored. The default is 20.
Returns
-------
Cell_Finded : (Dictionary)
All value will return to this dictionary. Keys are shown below:
['All_Cell_Information']: skimage data, contains all cell location and areas.
['Cell_Graph']:Cell location graph.
['Annotate_Cell_Graph']:Annotated cell graph, cell graph with number.
['Combine_Graph']:Circle cell on input graph.
['Combine_Graph_With_Number']:Circle Cell on input graph, and annotate cell number on input graph.
"""
# Get Original Cell Graph
Cell_Finded = {}
sharp_mask = Calculator.Normalized_2D_Gaussian_Generator(sharp_gauss)
back_mask = Calculator.Normalized_2D_Gaussian_Generator(back_gauss)
sharp_blur = scipy.ndimage.correlate(input_graph,sharp_mask,mode = 'reflect').astype('f8')
back_blur = scipy.ndimage.correlate(input_graph,back_mask,mode = 'reflect').astype('f8')
im_cell = (sharp_blur-back_blur)/np.max(sharp_blur-back_blur) # Local Max value shown in this graph.
if find_thres == 'otsu':# Add otsu method here.
level = filters.threshold_otsu(im_cell)
else:
level = np.mean(im_cell)+find_thres*np.std(im_cell)
origin_cells = im_cell>level # Original cell graphs, need to be filtered later.
# Revome boulder.
origin_cells = Graph_Tools.Boulder_Fill(origin_cells, shape_boulder, 0)
# Then remove small areas, other removal have no direct function, so we have to do it later.
cell_washed = skimage.morphology.remove_small_objects(origin_cells,min_pix,connectivity = 1)
# Get Cell Group Description here.
cell_label = skimage.measure.label(cell_washed)
All_Cells = skimage.measure.regionprops(cell_label)
# Then, wash bigger and oversize cell
for i in range(len(All_Cells)-1,-1,-1): # opposite sequence to avoid id change of pop.
current_cell = All_Cells[i]
area = current_cell.convex_area # Cell areas
# wash too big cell
if area > max_pix:
All_Cells.pop(i)
# wash oversize cell
else:
size = np.shape(All_Cells[i].image)
if np.max(size)>size_limit:
All_Cells.pop(i)
# Till now, Cell find is done, Graph Show shall be done here.
Cell_Finded['All_Cell_Information'] = All_Cells
Cell_Finded['Cell_Graph'] = Visualize.Cell_Visualize(All_Cells)
Cell_Finded['Annotate_Cell_Graph'] = Visualize.Label_Cell(Cell_Finded['Cell_Graph'],All_Cells)
# Then draw cell on input graph.
circled_cell = Visualize.Cell_Visualize(All_Cells,color = [0,255,100],mode = 'Boulder')
input_8bit = Graph_Tools.Graph_Depth_Change(input_graph,output_bit = 'u1')
input_8bit = cv2.cvtColor(input_8bit,cv2.COLOR_GRAY2RGB)
Cell_Finded['Combine_Graph'] = Graph_Tools.Graph_Combine(input_8bit,circled_cell)
Cell_Finded['Combine_Graph_With_Number'] = Visualize.Label_Cell(Cell_Finded['Combine_Graph'], All_Cells)
return Cell_Finded