如何增强图像，然后使用Python和OpenCV将其转换为二进制图像？

Question

我创建了一个简单的例子来说明我正在努力做的事情：

#!/usr/bin/python

#---------------------------------------------------------------------
# imports
#    >>> cv2.__version__
#    '2.4.6.1'
#    >>> scipy.__version__
#    '0.12.0'
#---------------------------------------------------------------------
import cv2
from scipy.signal import wiener

#---------------------------------------------------------------------
# functions
#---------------------------------------------------------------------
def enhance_image(input_image):
    my_im = input_image.copy()
    my_im = wiener(my_im)
    return my_im

#---------------------------------------------------------------------
# Main
#---------------------------------------------------------------------
filename="./data/5b2013d0-7939-48a6-94eb-045495b85343.png"

im=cv2.imread(filename)
cv2.imshow('Original',im)

gray_im = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
cv2.imshow('Gray',gray_im)

enhanced_im=enhance_image(gray_im)
cv2.imshow('Enhanced',enhanced_im)

# this fails
(thresh, bw_im) = cv2.threshold(enhanced_im, 128, 255, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)
cv2.imshow('Black and White',bw_im)

key = cv2.waitKey()

代码加载图像，将其转换为灰度，并使用scipy。信号模块运行维纳算法来增强图像，之后我想将图像转换为二值图像（黑白）。当我运行上面的代码时，我得到以下错误：

OpenCV错误：在阈值文件/home/mrichey/Downloads/opencv2中的断言失败（src.type（）==CV_8UC1）。4/opencv-2.4.6.1/modules/imgproc/src/thresh。cpp，第719行回溯（最后一次调用）：文件“/simple\u example.py”，第37行，in（thresh，bw\u im）=cv2。阈值（增强的im，128255，cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU）cv2。错误：/home/mrichey/Downloads/opencv2。4/opencv-

我对这个错误的研究使我相信，传递到阈值函数中的图像是错误的类型，但我不知道它需要转换成什么或如何转换。

当我将gray\u im传递到threshold时，它工作正常。我注意到gray\u im都是整数，但是enhanced\u im有浮点值，所以我尝试修改我的enhanced\u image函数：

def enhance_image(input_image):
    my_im = input_image.copy()
    my_im = wiener(my_im)
    return my_im.astype(int)

然而，我也犯了同样的错误。

Answer 1

将enhanced_im转换为threshold的适当类型将使其正常工作。这可以通过以下方式实现：

enhanced_im  = numpy.array(enhanced_im, numpy.uint8)

这是您的代码，经过修改后可以使用：

import cv2
from scipy.signal import wiener
import numpy
#---------------------------------------------------------------------
# functions
#---------------------------------------------------------------------
def enhance_image(input_image):
    my_im = input_image.copy()
    my_im = wiener(my_im)
    return my_im

#---------------------------------------------------------------------
# Main
#---------------------------------------------------------------------
filename="/home/ryan/OpenCV/opencv-2.3.4.7/samples/cpp/baboon.jpg"

im=cv2.imread(filename)
cv2.imshow('Original',im)

gray_im = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
cv2.imshow('Gray',gray_im)

enhanced_im=enhance_image(gray_im)
cv2.imshow('Enhanced',enhanced_im)
#print type(enhanced_im)
#enhanced_im.astype(numpy.uint8)
enhanced_im  = numpy.array(enhanced_im, numpy.uint8)
#print enhanced_im
# this fails
#ret,thresh = cv2.threshold(imgray,self.t1,225, cv2.THRESH_BINARY)
thresh, bw_im = cv2.threshold(enhanced_im, 128, 255, cv2.THRESH_BINARY)
cv2.imshow('Black and White',bw_im)

key = cv2.waitKey()