質問編集履歴

プログラム全文を掲載しました

2021/01/19 10:22

投稿

vamrat

スコア20

test CHANGED Viewed

File without changes

test CHANGED Viewed

@@ -146,7 +146,225 @@
+```analyze.py
+import cv2
+import numpy as np
+import matplotlib.pyplot as plt
+import methods
+# 入力画像読込
+# 画像サイズ（500(width) x 218(height) pix）
+img = cv2.imread('images/madslide12.jpg', cv2.IMREAD_COLOR)
+# オリジナル画像保存
+# import pdb; pdb.set_trace()
+org = img.copy()
+cv2.imwrite('results/original.png', org)
+# 画像ファイル
+#   - 画像データを処理プログラムに送る
+methods.image(org,img)
+# PyMeanShift
+#   - 第１引数：探索範囲、第２引数：探索色相、第３引数：粗さ
+methods.meanshift(12,3,200)
+# ヒストグラム均一化
+methods.contrast()
+# 類似色統合
+methods.clustering()
+# ブロック分割
+methods.division()
+# カラーラベリング
+methods.labeling()
+```
-```python
+```methods.py
+import cv2
+import matplotlib.pyplot as plt
+import matplotlib as mpl
+import numpy as np
+import pymeanshift as pms
+import os
+import sys
+from PIL import Image
+# sys.setrecursionlimit(8000) # 200 x 113 pix
+sys.setrecursionlimit(30000)  # 500 x 281 pix
+plt.gray()
+def image(_org,_img):
+  global org,img,h,w,c
+  global bo,go,ro,al
+  org,img = _org,_img
+  h,w,c = img.shape
+  bo,go,ro = cv2.split(org)
+  al = 0.55
+def meanshift(spatial_radius,range_radius,min_density):
+  global img
+  (img,labels,num) = pms.segment(cv2.cvtColor(img,cv2.COLOR_BGR2Lab),spatial_radius,range_radius,min_density)
+  img = cv2.cvtColor(img, cv2.COLOR_Lab2BGR)
+  cv2.imwrite('results/meanshift.png',img)
+def contrast():
+  global img
+  hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
+  h,s,v = cv2.split(hsv)
+  clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(3, 3))
+  result = clahe.apply(v)
+  hsv = cv2.merge((h,s,result))
+  img = cv2.cvtColor(hsv, cv2.COLOR_HSV2BGR)
+  cv2.imwrite("results/contrast.png", img)
+def division():
+  global block
+  hs,ws,cnt = 113,200,1
+  if ((h>hs)&(w>ws)):
+    for j in range(0,h,hs):
+      for i in range(0,w,ws):
+        if (((j+hs)<h)&((i+ws)<w)):
+          div = np.zeros((hs,ws,c),dtype=int)
+          div[0:hs,0:ws] = img[j:j+hs,i:i+ws]
+        else:
+          if (((i+ws)>w)&((j+hs)>h)):
+            div = np.zeros((h-j,w-i,c),dtype=int)
+            div[0:h-j,0:w-i] = img[j:h,i:w]
+          elif ((j+hs)>h):
+            div = np.zeros((h-j,ws,c),dtype=int)
+            div[0:h-j,0:ws] = img[j:h,i:i+ws]
+          else:
+            div = np.zeros((hs,w-i,c),dtype=int)
+            div[0:hs,0:w-i] = img[j:j+hs,i:w]
+        cv2.imwrite('results/division/division{}.png'.format(cnt),div)
+        cnt += 1
+  block = cnt
+  print('block number :',block)
+def clustering():
+  global img
+  im = Image.open('results/contrast.png')
+  im_q = im.quantize(colors=128, method=0, dither=1)
+  im_q.save('results/clustering.png')
+  img = cv2.imread('results/clustering.png', cv2.IMREAD_COLOR)
 def approximation(pix1,pix2):