python numpy 显示图像阵列的实例

每次要显示图像阵列的时候，使用自带的 matplotlib 或者cv2 都要设置一大堆东西，subplot，fig等等，突然想起 可以利用numpy 的htstack() 和 vstack()将图片对接起来组成一张新的图片。因此写了写了下面的函数。做了部分注释，一些比较绕的地方可以自行体会。

大致流程包括：

1、输入图像列表img_list

2、show_type : 最终的显示方式，输入为行数列数（例如show_type=22 ，则最终显示图片为两行两列）

3、basic_shape, 图片ｒｅｓｉｚｅ的尺寸。

def image_show( img_list, show_type, basic_size=[300,500]):
'''
 img_list contains the images that need to be stitched,
 the show_typ contains the final shape of the stitched one, ie, 12 for 1 row 2 cols.
 basic_size : all input image need to be reshaped first.

'''
# reshap row and col number.
n_row, n_col = basic_size
#print n_row,n_col

# num of pixels need to be filled vertically and horizontally.
h_filling = 10
v_filling = 10

# image resize.
resize_list=[]
for i in img_list:
 temp_img = cv2.resize( i, ( n_col, n_row ), interpolation = cv2. INTER_CUBIC )
 resize_list.append( temp_img )

# resolve the final stitched image 's shape.
n_row_img, n_col_img = show_type/10, show_type％10
#print n_row_img, n_col_img

# the blank_img and the image need to be filled should be defined firstly.
blank_img= np.ones([n_row,n_col])*255
blank_img= np.array( blank_img, np.uint8 )
v_img= np.array( np.ones([n_row,v_filling])*255, np.uint8)
h_img= np.array( np.ones ([ h_filling, n_col_img*n_col+(n_col_img-1)*h_filling])*255, np.uint8)

# images in the image list should be dispatched into different sub-list
# in each sub list the images will be connected horizontally.
recombination_list=[]
temp_list=[]
n_list= len(resize_list)
for index, i in enumerate ( xrange (n_list)):
 if index!= 0 and index ％ n_col_img==0 :
  recombination_list.append(temp_list)
  temp_list = []
  if len(resize_list)> n_col_img:
   pass
  else:
   recombination_list.append(resize_list)
   break
 temp_list.append( resize_list.pop(0))
if n_list== n_col_img:
 recombination_list.append(temp_list)
#print len(temp_list)
#print temp_list

# stack the images horizontally.
h_temp=[]
for i in recombination_list:
 #print len(i)
 if len(i)==n_col_img:

temp_new_i=[ [j,v_img] if index+1 != len(i) else j for index, j in enumerate (i) ]
  new_i=[ j for i in temp_new_i[:-1] for j in i ]
  new_i.append( temp_new_i[-1])
  h_temp.append(np.hstack(new_i))
 else:

add_n= n_col_img - len(i)
  for k in range(add_n):
   i.append(blank_img)

temp_new_i=[ [j,v_img] if index+1 != len(i) else j for index, j in enumerate (i) ]
  new_i=[ j for i in temp_new_i[:-1] for j in i ]
  new_i.append( temp_new_i[-1])

h_temp.append(np.hstack(new_i))

#print len(h_temp)
#print h_temp

temp_full_img= [ [j, h_img ] if index+1 != len(h_temp) else j for index, j in enumerate(h_temp) ]
if len(temp_full_img) > 2:
 full_img= [ j for i in temp_full_img[:-1] for j in i ]
 full_img.append(temp_full_img[-1])
else:
 full_img= [ j for i in temp_full_img for j in i ]
 #full_img.append(temp_full_img[-1])

if len(full_img)>1:
 return np.vstack( full_img)
else:
 return full_img

最终输入情况和结果如下图：

第一组结果图：自行看输入

第二组结果图。

来源：https://blog.csdn.net/vola9527/article/details/52801380