def hist_compare(image1, image2):
size = image1.shape[:2]
image1 = cv.resize(image1, size)
image2 = cv.resize(image2, size)
image1 = cv.cvtColor(image1, cv.COLOR_BGR2GRAY)
image2 = cv.cvtColor(image2, cv.COLOR_BGR2GRAY)
# image1 = cv.equalizeHist(image1)
# image2 = cv.equalizeHist(image2)
hist1 = cv.calcHist([image1], [0], None, [256], [0, 256])
hist2 = cv.calcHist([image2], [0], None, [256], [0, 256])
match1 = cv.compareHist(hist1, hist2, cv.HISTCMP_BHATTACHARYYA)
match2 = cv.compareHist(hist1, hist2, cv.HISTCMP_CORREL)
match3 = cv.compareHist(hist1, hist2, cv.HISTCMP_CHISQR)
# 巴氏距离越小越相似，相关性越大越相似，卡方越大越不相似
print("巴氏距离", match1)
print("相关性", match2)
print("卡方", match3)
# 直方图绘制
# 2-2-1
plt.subplot(221)
plt.imshow(image1, cmap=plt.cm.gray)
# 2-2-3
plt.subplot(223)
plt.imshow(image2, cmap=plt.cm.gray)
# 2-2-2
plt.subplot(222)
plt.plot(hist1) # 线
# 2-2-4
plt.subplot(224)
plt.plot(hist2) # 线
plt.show()
def hist_compare(image1, image2):
hist1 = creat_rgb_hist(image1)
hist2 = creat_rgb_hist(image2)
match1 = cv.compareHist(hist1, hist2,
cv.HISTCMP_BHATTACHARYYA) #巴氏距离，越小越相似
match2 = cv.compareHist(hist1, hist2, cv.HISTCMP_CORREL) #相关性比较，越大越相似
match3 = cv.compareHist(hist1, hist2, cv.HISTCMP_CHISQR) #卡方，越大越不相似
print("巴氏：%s,相关性：%s,卡方：%s" % (match1, match2, match3))
def calc_difference(hist_1, hist_2):
img_hist_diff = cv2.compareHist(hist_1, hist_2, cv2.HISTCMP_BHATTACHARYYA)
img_template_probability_match = cv2.matchTemplate(
hist_1, hist_2, cv2.TM_CCOEFF_NORMED)[0][0]
img_template_diff = 1 - img_template_probability_match
# taking only 10% of histogram diff, since it's less accurate than template method
commutative_image_diff = (img_hist_diff / 10) + img_template_diff
logging.debug("calculated image difference %f", commutative_image_diff)
return commutative_image_diff
def hist_diff(img_1, img_2):
""" Calculates distance between images based on the distance between their
histograms. Example found at:
https://www.pyimagesearch.com/2014/07/14/3-ways-compare-histograms-using-opencv-python/
"""
hist_1 = cv2.calcHist([img_1], [0, 1, 2], None, [8, 8, 8],
[0, 256, 0, 256, 0, 256])
hist_1 = cv2.normalize(hist_1, hist_1).flatten()
hist_2 = cv2.calcHist([img_2], [0, 1, 2], None, [8, 8, 8],
[0, 256, 0, 256, 0, 256])
hist_2 = cv2.normalize(hist_2, hist_2).flatten()
return cv2.compareHist(hist_1, hist_2, cv2.HISTCMP_BHATTACHARYYA)
def compare(img_source, img_reference):
hist_requete = cv.calcHist([img_requete], [0], None, [256], [0, 256])
hist_ref = cv.calcHist([img_reference], [0], None, [256], [0, 256])
current_matching_value = cv.compareHist(hist_requete, hist_ref,
cv.HISTCMP_CORREL)
return current_matching_value
break
for r_x in range(0, width, slide):
sys.stdout.write("\r load_n : %d" % read_num)
sys.stdout.flush()
if r_x + size >= width:
break
dst = src[r_y:r_y + size, r_x:r_x + size] #image
resize_img = cv2.resize(dst, (resize_shape))
X.append(resize_img) #image_list
r.append([r_y, r_x]) #coordinate_list
resize_img = resize_img * 255.0
img_hist = cv2.calcHist([resize_img], [0], None, [256], [0, 256])
ret = cv2.compareHist(white_hist, img_hist, 0)
ret_list.append(ret) #background_list
read_num += 1
X = np.asarray(X)
predicted_value = model.predict(X, verbose=0)
for nth, r_n in enumerate(r):
if ret_list[nth] < 1.0:
predicted_keeper[r_n[0]:r_n[0] + size,
r_n[1]:r_n[1] + size] += predicted_value[nth]
num_inferences[r_n[0]:r_n[0] + size, r_n[1]:r_n[1] + size] += 1
for colormap_y in range(height):
for colormap_x in range(width):
def classifyImageByHist(imagePath, datasetHists, showStep=False):
imageName = os.path.basename(imagePath)
imageBGR = cv2.imread(imagePath)
imageFeatureVector = imageBGR2FeatureVector(imageBGR)
imageHist = imageBGR2FlatHist(imageBGR)
result = []
resultLastIndex = 0
if showStep:
for key, item in datasetHists.items():
for i in range(len(datasetHists[key])):
imageHistFormatted = []
print("\n" + key + "[" + str(i) + "]: ")
for j in range(len(datasetHists[key][i])):
imageHistFormatted.append('{0:.5f}'.format(
datasetHists[key][i][j]))
for n in range(len(imageHistFormatted)):
# if n % 8 == 0:
# print('')
print(imageHistFormatted[n], end=' ')
if n == len(imageHistFormatted) - 1:
print()
print("\nuji:")
for n in range(len(imageHist)):
print('{0:.5f}'.format(imageHist[n]), end=' ')
if n == len(imageHist) - 1:
print()
for key, item in datasetHists.items():
for i in range(len(datasetHists[key])):
imageHistFormatted = []
print("\nkomparasi uji dengan " + key + "[" + str(i) + "]: ")
for j in range(len(datasetHists[key][i])):
buffer = min(imageHist[j], datasetHists[key][i][j])
imageHistFormatted.append('{0:.5f}'.format(buffer))
for n in range(len(imageHistFormatted)):
print(imageHistFormatted[n], end=' ')
if n == len(imageHistFormatted) - 1:
print()
imageHistFormatted = np.array(imageHistFormatted).astype(
np.float)
result.append((key, sum(imageHistFormatted)))
# print("hasil sum: %f" % (result[resultLastIndex][1]))
resultLastIndex += 1
else:
for key, value in datasetHists.items():
for i in range(6):
imageHistDataset = value[i]
histValue = '{0:.5f}'.format(
cv2.compareHist(imageHist, imageHistDataset,
cv2.HISTCMP_INTERSECT))
result.append((key, float(histValue)))
# print(str(key) + "[" + str(i) + "]: " + str(result[resultLastIndex][1]))
resultLastIndex += 1
print(
"[ K O M P A R A S I G A M B A R %s D E N G A N D A T A S E T ]"
% (imageName))
result.sort(key=lambda tup: tup[1], reverse=True)
for r in result:
print(r)
classes = []
for key, value in datasetHists.items():
classes.append((key, 0))
k = 5
print()
print("[ P E M B E R I A N R A N K I N G D E N G A N K = %d ]" %
(k))
for i in range(len(classes)):
for j in range(k):
if result[j][0] == classes[i][0]:
classes[i] = (classes[i][0], classes[i][1] + 1)
classes.sort(key=lambda tup: tup[1], reverse=True)
for c in classes:
print(c)
print()
print("[ H A S I L K L A S I F I K A S I ]")
print("Kelas gambar %s adalah %s" % (imageName, classes[0][0]))
print(
"________________________________________________________________________________\n"
)
def get_match_rate(source, area):
"""" Retrieve a similitude rate between a source image and a template """
hist_source = cv.calcHist([source], [0], None, [256], [0, 256])
hist_area = cv.calcHist([area], [0], None, [256], [0, 256])
return cv.compareHist(hist_source, hist_area, cv.HISTCMP_CORREL)