OpenCV threshold函数详解
threshold函数作用：去掉噪，例如过滤很小或很大像素值的图像点。threshold函数原型：double threshold( InputArray src, OutputArray dst,double thresh, double maxval, int type );参数说明：src：原图像。dst：结果图像。thresh...
文章共2,575字 · 阅读需要大约9分钟
一键AI生成摘要，助你高效阅读
问答
·
threshold函数作用：
去掉噪，例如过滤很小或很大像素值的图像点。
threshold函数python版原型：
retval, dst = cv.threshold( src, thresh, maxval, type[, dst] ) 参数说明：
src：原图像。
dst：结果图像。
thresh：当前阈值。
maxVal：最大阈值，一般为255.
thresholdType:阈值类型，主要有下面几种：
enum ThresholdTypes {
THRESH_BINARY = 0,
THRESH_BINARY_INV = 1,
THRESH_TRUNC = 2,
THRESH_TOZERO = 3,
THRESH_TOZERO_INV = 4,
THRESH_MASK = 7,
THRESH_OTSU = 8,
THRESH_TRIANGLE = 16
};
返回值:
retval： 与参数thresh一致
dst： 结果图像
| 编号 | 阈值类型枚举 |
注意
|
| 1 | THRESH_BINARY | |
| 2 | THRESH_BINARY_INV | |
| 3 | THRESH_TRUNC | |
| 4 | THRESH_TOZERO | |
| 5 | THRESH_TOZERO_INV | |
| 6 | THRESH_MASK |
不支持
|
| 7 | THRESH_OTSU |
不支持32位
|
| 8 | THRESH_TRIANGLE |
不支持32位
|
具体说明如下：
注意：
THRESH_OTSU和THRESH_TRIANGLE是作为优化算法配合THRESH_BINARY、THRESH_BINARY_INV、THRESH_TRUNC、THRESH_TOZERO以及THRESH_TOZERO_INV来使用的。
当使用了THRESH_OTSU和THRESH_TRIANGLE两个标志时，输入图像必须为单通道。
生成关系：
下面通过代码来验证一下threshold的用法：
import numpy as np
import cv2
#创建一个6x6的BGR三通道图片
img6x6 = np.zeros((6,6,3), dtype=np.uint8)
#Blue通道取值0~99随机值
img6x6[:,:,0] = np.random.randint(0, 100, size=36).reshape(6,6)
#Green通道取值100~199随机值
img6x6[:,:,1] = np.random.randint(100, 200, size=36).reshape(6,6)
#Red通道取值200~255随机值
img6x6[:,:,2] = np.random.randint(200, 256, size=36).reshape(6,6)
print(img6x6)得到img6x6的BGR三通道数据：
[[[ 62 171 226]
[ 29 118 212]
[ 61 151 200]
[ 43 123 206]
[ 58 146 232]
[ 2 124 246]]
[[ 48 101 207]
[ 32 194 228]
[ 90 171 241]
[ 36 152 244]
[ 64 169 245]
[ 71 106 202]]
[[ 27 198 206]
[ 39 193 213]
[ 60 128 207]
[ 53 134 248]
[ 25 187 229]
[ 72 116 229]]
[[ 52 163 201]
[ 39 189 251]
[ 63 190 217]
[ 73 139 212]
[ 54 166 236]
[ 43 191 200]]
[[ 34 119 207]
[ 36 110 209]
[ 40 191 229]
[ 60 100 228]
[ 62 174 250]
[ 41 115 209]]
[[ 35 165 246]
[ 50 182 218]
[ 77 110 239]
[ 44 106 245]
[ 21 189 252]
[ 64 144 241]]]128 207]
[ 53 134 248]
[ 25 187 229]
[ 72 116 229]]
[[ 52 163 201]
[ 39 189 251]
[ 63 190 217]
[ 73 139 212]
[ 54 166 236]
[ 43 191 200]]
[[ 34 119 207]
[ 36 110 209]
[ 40 191 229]
[ 60 100 228]
[ 62 174 250]
[ 41 115 209]]
[[ 35 165 246]
[ 50 182 218]
[ 77 110 239]
[ 44 106 245]
[ 21 189 252]
[ 64 144 241]]]观察结果我们发现img6x6[2][2][1]=128，我们取该值作为thresh值来演示THRESH_BINARY、THRESH_BINARY_INV等的作用。
1) THRESH_BINARY
我们取参数值thresh=128，maxval=255，flag = THRESH_BINARY
cv2.threshold(img6x6, 128, 255, cv2.THRESH_BINARY)根据前面对THRESH_BINARY flag值的解释：
预期得到的结果中，img6x6中>128的值均应变为255，<=128的值均应变为0。
我们实际得到的结果为：
[[[ 0 255 255]
[ 0 0 255]
[ 0 255 255]
[ 0 0 255]
[ 0 255 255]
[ 0 0 255]]
[[ 0 0 255]
[ 0 255 255]
[ 0 255 255]
[ 0 255 255]
[ 0 255 255]
[ 0 0 255]]
[[ 0 255 255]
[ 0 255 255]
[ 0 0 255]
[ 0 255 255]
[ 0 255 255]
[ 0 0 255]]
[[ 0 255 255]
[ 0 255 255]
[ 0 255 255]
[ 0 255 255]
[ 0 255 255]
[ 0 255 255]]
[[ 0 0 255]
[ 0 0 255]
[ 0 255 255]
[ 0 0 255]
[ 0 255 255]
[ 0 0 255]]
[[ 0 255 255]
[ 0 255 255]
[ 0 0 255]
[ 0 0 255]
[ 0 255 255]
[ 0 255 255]]]0 255]
[ 0 255 255]
[ 0 255 255]
[ 0 0 255]]
[[ 0 255 255]
[ 0 255 255]
[ 0 255 255]
[ 0 255 255]
[ 0 255 255]
[ 0 255 255]]
[[ 0 0 255]
[ 0 0 255]
[ 0 255 255]
[ 0 0 255]
[ 0 255 255]
[ 0 0 255]]
[[ 0 255 255]
[ 0 255 255]
[ 0 0 255]
[ 0 0 255]
[ 0 255 255]
[ 0 255 255]]]与预期的结果一致。
2）THRESH_BINARY_INV
THRESH_BINARY_INV的作用与THRESH_BINARY相反，大于thresh的值置0，小于等于thresh的值置maxval:
预期的结果与1）的结果完全相反。
code:
ret,dst=cv2.threshold(img6x6, 128, 255, cv2.THRESH_BINARY_INV)
print(dst)dst输出：
[[[255 0 0]
[255 255 0]
[255 0 0]
[255 255 0]
[255 0 0]
[255 255 0]]
[[255 255 0]
[255 0 0]
[255 0 0]
[255 0 0]
[255 0 0]
[255 255 0]]
[[255 0 0]
[255 0 0]
[255 255 0]
[255 0 0]
[255 0 0]
[255 255 0]]
[[255 0 0]
[255 0 0]
[255 0 0]
[255 0 0]
[255 0 0]
[255 0 0]]
[[255 255 0]
[255 255 0]
[255 0 0]
[255 255 0]
[255 0 0]
[255 255 0]]
[[255 0 0]
[255 0 0]
[255 255 0]
[255 255 0]
[255 0 0]
[255 0 0]]]
255 0]
[255 0 0]
[255 0 0]
[255 255 0]]
[[255 0 0]
[255 0 0]
[255 0 0]
[255 0 0]
[255 0 0]
[255 0 0]]
[[255 255 0]
[255 255 0]
[255 0 0]
[255 255 0]
[255 0 0]
[255 255 0]]
[[255 0 0]
[255 0 0]
[255 255 0]
[255 255 0]
[255 0 0]
[255 0 0]]]
与预期结果完全一致。
3）THRESH_TRUNC
THRESH_TRUNC的作用是对大于thresh的值进行截断，其余值保留原值。
code:
ret,dst=cv2.threshold(img6x6, 128, 255, cv2.THRESH_TRUNC)
print(dst)dst输出：
[[[ 62 128 128]
[ 29 118 128]
[ 61 128 128]
[ 43 123 128]
[ 58 128 128]
[ 2 124 128]]
[[ 48 101 128]
[ 32 128 128]
[ 90 128 128]
[ 36 128 128]
[ 64 128 128]
[ 71 106 128]]
[[ 27 128 128]
[ 39 128 128]
[ 60 128 128]
[ 53 128 128]
[ 25 128 128]
[ 72 116 128]]
[[ 52 128 128]
[ 39 128 128]
[ 63 128 128]
[ 73 128 128]
[ 54 128 128]
[ 43 128 128]]
[[ 34 119 128]
[ 36 110 128]
[ 40 128 128]
[ 60 100 128]
[ 62 128 128]
[ 41 115 128]]
[[ 35 128 128]
[ 50 128 128]
[ 77 110 128]
[ 44 106 128]
[ 21 128 128]
[ 64 128 128]]]128 128]
[ 60 128 128]
[ 53 128 128]
[ 25 128 128]
[ 72 116 128]]
[[ 52 128 128]
[ 39 128 128]
[ 63 128 128]
[ 73 128 128]
[ 54 128 128]
[ 43 128 128]]
[[ 34 119 128]
[ 36 110 128]
[ 40 128 128]
[ 60 100 128]
[ 62 128 128]
[ 41 115 128]]
[[ 35 128 128]
[ 50 128 128]
[ 77 110 128]
[ 44 106 128]
[ 21 128 128]
[ 64 128 128]]]4） THRESH_TOZERO
THRESH_TOZERO的作用是对大于thresh的值进行保留原值，小于或等于thresh的值置0：
code:
ret, dst=cv2.threshold(img6x6, 128, 255, cv2.THRESH_TOZERO)
print(dst)dst输出：
[[[ 0 171 226]
[ 0 0 212]
[ 0 151 200]
[ 0 0 206]
[ 0 146 232]
[ 0 0 246]]
[[ 0 0 207]
[ 0 194 228]
[ 0 171 241]
[ 0 152 244]
[ 0 169 245]
[ 0 0 202]]
[[ 0 198 206]
[ 0 193 213]
[ 0 0 207]
[ 0 134 248]
[ 0 187 229]
[ 0 0 229]]
[[ 0 163 201]
[ 0 189 251]
[ 0 190 217]
[ 0 139 212]
[ 0 166 236]
[ 0 191 200]]
[[ 0 0 207]
[ 0 0 209]
[ 0 191 229]
[ 0 0 228]
[ 0 174 250]
[ 0 0 209]]
[[ 0 165 246]
[ 0 182 218]
[ 0 0 239]
[ 0 0 245]
[ 0 189 252]
[ 0 144 241]]]
0 207]
[ 0 134 248]
[ 0 187 229]
[ 0 0 229]]
[[ 0 163 201]
[ 0 189 251]
[ 0 190 217]
[ 0 139 212]
[ 0 166 236]
[ 0 191 200]]
[[ 0 0 207]
[ 0 0 209]
[ 0 191 229]
[ 0 0 228]
[ 0 174 250]
[ 0 0 209]]
[[ 0 165 246]
[ 0 182 218]
[ 0 0 239]
[ 0 0 245]
[ 0 189 252]
[ 0 144 241]]]
5) THRESH_TOZERO_INV
THRESH_TOZERO_INV的作用与THRESH_TOZERO的作用完全相反，对大于thresh的值置0，其余值则保留：
code:
ret, dst=cv2.threshold(img6x6, 131, 255, cv2.THRESH_TOZERO_INV)
print(dst)dst输出：
[[[ 62 0 0]
[ 29 118 0]
[ 61 0 0]
[ 43 123 0]
[ 58 0 0]
[ 2 124 0]]
[[ 48 101 0]
[ 32 0 0]
[ 90 0 0]
[ 36 0 0]
[ 64 0 0]
[ 71 106 0]]
[[ 27 0 0]
[ 39 0 0]
[ 60 128 0]
[ 53 0 0]
[ 25 0 0]
[ 72 116 0]]
[[ 52 0 0]
[ 39 0 0]
[ 63 0 0]
[ 73 0 0]
[ 54 0 0]
[ 43 0 0]]
[[ 34 119 0]
[ 36 110 0]
[ 40 0 0]
[ 60 100 0]
[ 62 0 0]
[ 41 115 0]]
[[ 35 0 0]
[ 50 0 0]
[ 77 110 0]
[ 44 106 0]
[ 21 0 0]
[ 64 0 0]]]
128 0]
[ 53 0 0]
[ 25 0 0]
[ 72 116 0]]
[[ 52 0 0]
[ 39 0 0]
[ 63 0 0]
[ 73 0 0]
[ 54 0 0]
[ 43 0 0]]
[[ 34 119 0]
[ 36 110 0]
[ 40 0 0]
[ 60 100 0]
[ 62 0 0]
[ 41 115 0]]
[[ 35 0 0]
[ 50 0 0]
[ 77 110 0]
[ 44 106 0]
[ 21 0 0]
[ 64 0 0]]]
6）THRESH_OTSU
THRESH_OTSU是需要搭配THRESH_BINARY、THRESH_BINARY_INV、THRESH_TRUNC、THRESH_TOZERO以及THRESH_TOZERO_INV来使用的。并且输入图像必须是灰度图像。
所以，我们首先需要创建一个灰度图像：
code:
img_gray = cv2.cvtColor(img6x6,cv2.COLOR_BGR2GRAY)
print(img_gray)得到的img_gray为：
[[175 0 0 0 162 0]
[ 0 186 183 166 180 0]
[181 181 0 159 181 0]
[162 190 184 0 174 177]
[ 0 0 185 0 184 0]
[174 178 0 0 189 164]]
然后在通过THRESH_OTSU标准搭配其他标志位来对输入图像做处理，这里只演示搭配THRESH_BINARY标志的用法，
code:
ret,dst=cv2.threshold(img_gray, 127, 255, cv2.THRESH_TOZERO|cv2.THRESH_OTSU)
print(dst)dst输出：
[[175 0 0 0 162 0]
[ 0 186 183 166 180 0]
[181 181 0 159 181 0]
[162 190 184 0 174 177]
[ 0 0 185 0 184 0]
[174 178 0 0 189 164]]
对比一下不使用THRESH_OTSU时的输出，
code:
ret,dst=cv2.threshold(img_gray, 166, 255, cv2.THRESH_TOZERO)
print(dst)dst输出:
[[175 0 0 0 0 0]
[ 0 186 183 0 180 0]
[181 181 0 0 181 0]
[ 0 190 184 0 174 177]
[ 0 0 185 0 184 0]
[174 178 0 0 189 0]]0 180 0]
[181 181 0 0 181 0]
[ 0 190 184 0 174 177]
[ 0 0 185 0 184 0]
[174 178 0 0 189 0]]7) THRESH_TRIANGLE
THRESH_TRIANGLE用法与THRESH_OTSU用法一致，这里不做演示。
综合示例
Demo Code
#encoding=utf-8
import sys, cv2, numpy as np
from ui_thresholdDemo import Ui_ThresoldDemo
from PyQt5.QtWidgets import QApplication, QWidget, QFileDialog
from PyQt5.QtGui import QImage, QPixmap
class ThresholdDemo(QWidget, Ui_ThresoldDemo):
def __init__(self):
super().__init__()
self.setupUi(self)
self.setupSigSlots()
self.param_threshold = 1
self.param_maxVal = 255
self.param_thresholdType = cv2.THRESH_BINARY
self.param_ExtendType = 0
self.checkGray = False
def setupSigSlots(self):
self.btnBrowse.clicked.connect(self.slotBtnBrowse)
self.parmThreshold.valueChanged.connect(self.slotThresholdChanged)
self.paramMaxVal.valueChanged.connect(self.slotMaxValChanged)
self.parmThresholdType.currentTextChanged[str].connect(self.slotThresholdTypeChanged)
self.btnCheckGray.toggled.connect(self.slotCheckGray)
self.parmThresholdType2.currentTextChanged[str].connect(self.slotThresholdType2Changed)
def slotBtnBrowse(self):
try:
file, _ = QFileDialog.getOpenFileName(self, 'Open Image', 'D:\\tmp\\pics', '*.*', '*.*',)
if file:
self.imagePath.setText(file)
img_init = cv2.imdecode(np.fromfile(file, dtype=np.uint8), -1)
self.img_threshold = img_init.copy()
cv2.cvtColor(img_init, cv2.COLOR_BGR2RGB, img_init)
h,w,c = img_init.shape
qImg = QImage(img_init, w, h, w*c, QImage.Format_RGB888)
self.initPicFrame.setPixmap(QPixmap.fromImage(qImg))
self.updateThresholdImage()
except Exception as e:
print(e)
def updateThresholdImage(self):
try:
img = self.img_threshold.copy()
if self.checkGray:
img_gray = np.zeros((img.shape[0], img.shape[1]), dtype=np.uint8)
cv2.cvtColor(img, cv2.COLOR_BGR2GRAY, img_gray)
ret, img_threshold = cv2.threshold(img_gray, self.param_threshold, self.param_maxVal, self.param_thresholdType|self.param_ExtendType)
print('thresh:{}, maxval:{}, flags:{}|{}'.format(self.param_threshold, self.param_maxVal, self.param_thresholdType, self.param_ExtendType))
h,w = img_threshold.shape
qImg = QImage(img_threshold, w, h, w, QImage.Format_Indexed8)
self.thresholdPicFrame.setPixmap(QPixmap.fromImage(qImg))
else:
ret, img_threshold = cv2.threshold(img, self.param_threshold, self.param_maxVal, self.param_thresholdType)
#ret: ret equals to self.param_threshold, it canbe ignored
h,w,c = img_threshold.shape
cv2.cvtColor(img_threshold, cv2.COLOR_BGR2RGB, img_threshold)
qImg = QImage(img_threshold, w, h, w*c, QImage.Format_RGB888)
self.thresholdPicFrame.setPixmap(QPixmap.fromImage(qImg))
except Exception as e:
print(e)
def slotThresholdChanged(self, val):
self.param_threshold = val
self.labelThreshold.setText('threshold:{:>3}'.format(val))
self.updateThresholdImage()
def slotMaxValChanged(self, val):
self.param_maxVal = val
self.labelMaxVal.setText(' maxValue:{:>3}'.format(val))
self.updateThresholdImage()
def slotThresholdTypeChanged(self, type):
type_dic = {
'THRESH_BINARY': cv2.THRESH_BINARY,
'THRESH_BINARY_INV': cv2.THRESH_BINARY_INV,
'THRESH_TRUNC': cv2.THRESH_TRUNC,
'THRESH_TOZERO': cv2.THRESH_TOZERO,
'THRESH_TOZERO_INV': cv2.THRESH_TOZERO_INV,
#'THRESH_MASK': cv2.THRESH_MASK,
#'THRESH_OTSU': cv2.THRESH_OTSU,
#'THRESH_TRIANGLE': cv2.THRESH_TRIANGLE,
}
self.param_thresholdType = type_dic[type]
#print(type,self.param_thresholdType )
self.updateThresholdImage()
def slotThresholdType2Changed(self, type):
try:
extend_type = {
"THRESH_OTSU":cv2.THRESH_OTSU,
"THRESH_TRIANGLE":cv2.THRESH_TRIANGLE
}
if type == 'NONE':
self.param_ExtendType = 0
self.checkGray = False
self.btnCheckGray.setEnabled(True)
else:
self.param_ExtendType = extend_type[type]
self.checkGray = True
self.btnCheckGray.setEnabled(False)
print('self.param_ExtendType: ', self.param_ExtendType)
self.updateThresholdImage()
except Exception as e:
print(e)
def slotCheckGray(self, bGray):
self.checkGray = bGray
self.updateThresholdImage()
def thresholdDemo():
img6x6 = np.zeros((6,6,3), dtype=np.uint8)
img6x6[:,:,0] = np.random.randint(0, 100, size=36).reshape(6,6)
img6x6[:,:,1] = np.random.randint(100, 200, size=36).reshape(6,6)
img6x6[:,:,2] = np.random.randint(200, 256, size=36).reshape(6,6)
cv2.threshold(img6x6, 127, 255, cv2.THRESH_BINARY)
if __name__ == '__main__':
qApp = QApplication(sys.argv)
demo = ThresholdDemo()
demo.show()
qApp.exec_()
程序运行结果：
完整Code下载，请点这里
旨在为数千万中国开发者提供一个无缝且高效的云端环境，以支持学习、使用和贡献开源项目。
更多推荐
48
0- 0
已为社区贡献1条内容
所有评论(0)