基于Python与OpenCV实现目录图片去重：感知哈希算法实战
本文介绍利用Python和OpenCV库，通过感知哈希算法对指定目录下的图片进行自动去重。详细阐述了环境配置、算法原理（缩放、灰度化、指纹生成及汉明距离计算），并提供了完整的Python脚本与递归遍历的Shell脚本示例，帮助开发者高效清理重复图片。
版本信息
运行平台为Ubuntu 14系统，硬件配置为Intel I5处理器及4GB内存。
Python解释器版本指定为python2.7。
OpenCV库版本采用2.13.4。
环境依赖配置
若系统中未预装Python，需先执行安装操作。
执行命令：sudo apt-get install python
执行命令：sudo apt-get install python-dev
执行命令：sudo apt-get install python-pip
执行命令：sudo pip install numpy mathplotlib
执行命令：sudo apt-get install libcv-dev
执行命令：sudo apt-get install python-opencv
采用感知哈希算法执行图片去重操作。
算法逻辑是对目录内所有文件进行遍历比对，因此图片数量越多处理速度越慢，但能显著减少手动筛选的工作量。
感知哈希算法的具体原理如下：
第一步，将所有待比较的图片统一缩放至8*8像素的灰度图像。
第二步，计算每个像素值与图像平均灰度值的差异，从而生成代表图片特征的指纹序列。
第三步，基于生成的指纹序列计算汉明距离。
第四步，若计算出的不同像素点数量小于5，则判定为相同或高度相似的图片。
#!/usr/bin/python # -*- coding: UTF-8 -*- import cv2 import numpy as np import os,sys,types
def cmpandremove2(path):
dirs = os.listdir(path)
dirs.sort()
if len(dirs) <= 0:
return
dict={}
for i in dirs:
prepath = path + "/" + i
preimg = cv2.imread(prepath)
if type(preimg) is types.NoneType:
continue
preresize = cv2.resize(preimg, (8,8))
pregray = cv2.cvtColor(preresize, cv2.COLOR_BGR2GRAY)
premean = cv2.mean(pregray)[0]
prearr = np.array(pregray.data)
for j in range(0,len(prearr)):
if prearr[j] >= premean:
prearr[j] = 1
else:
prearr[j] = 0
print "get", prepath
dict[i] = prearr
dictkeys = dict.keys()
dictkeys.sort()
index = 0
while True:
if index >= len(dictkeys):
break
curkey = dictkeys[index]
dellist=[]
print curkey
index2 = index
while True:
if index2 >= len(dictkeys):
break
j = dictkeys[index2]
if curkey == j:
index2 = index2 + 1
continue
arr1 = dict[curkey]
arr2 = dict[j]
diff = 0
for k in range(0,len(arr2)):
if arr1[k] != arr2[k]:
diff = diff + 1
if diff <= 5:
dellist.append(j)
index2 = index2 + 1
if len(dellist) > 0:
for j in dellist:
file = path + "/" + j
print "remove", file
os.remove(file)
dict.pop(j)
dictkeys = dict.keys()
dictkeys.sort()
index = index + 1
def cmpandremove(path):
index = 0
flag = 0
dirs = os.listdir(path)
dirs.sort()
if len(dirs) <= 0:
return 0
while True:
if index >= len(dirs):
break
prepath = path + dirs[index]
print prepath
index2 = 0
preimg = cv2.imread(prepath)
if type(preimg) is types.NoneType:
index = index + 1
continue
preresize = cv2.resize(preimg,(8,8))
pregray = cv2.cvtColor(preresize, cv2.COLOR_BGR2GRAY)
premean = cv2.mean(pregray)[0]
prearr = np.array(pregray.data)
for i in range(0,len(prearr)):
if prearr[i] >= premean:
prearr[i] = 1
else:
prearr[i] = 0
removepath = []
while True:
if index2 >= len(dirs):
break
if index2 != index:
curpath = path + dirs[index2]
#print curpath
curimg = cv2.imread(curpath)
if type(curimg) is types.NoneType:
index2 = index2 + 1
continue
curresize = cv2.resize(curimg, (8,8))
curgray = cv2.cvtColor(curresize, cv2.COLOR_BGR2GRAY)
curmean = cv2.mean(curgray)[0]
curarr = np.array(curgray.data)
for i in range(0,len(curarr)):
if curarr[i] >= curmean:
curarr[i] = 1
else:
curarr[i] = 0
diff = 0
for i in range(0,len(curarr)):
if curarr[i] != prearr[i] :
diff = diff + 1
if diff <= 5:
print 'the same'
removepath.append(curpath)
flag = 1
index2 = index2 + 1
index = index + 1
if len(removepath) > 0:
for file in removepath:
print "remove", file
os.remove(file)
dirs = os.listdir(path)
dirs.sort()
if len(dirs) <= 0:
return 0
#index = 0
return flag
def main(argv):
if len(argv) <= 1:
print "command error"
return -1
if os.path.exists(argv[1]) is False:
return -1
path = argv[1]
'''
while True:
if cmpandremove(path) == 0:
break
'''
cmpandremove(path)
return 0
if __name__ == '__main__':
main(sys.argv)
为简化操作流程，脚本会遍历指定目录及其子目录，对需要去重的文件夹执行批量处理。
#!/bin/bash
indir=$1
addcount=0
function intest()
{
for file in $1/*
do
echo $file
if test -d $file
then
~/similar.py $file/
intest $file
fi
done
}
intest $indir
以上即为使用Python和OpenCV对目录下图片进行去重的完整方法分享，供读者参考与实践，欢迎支持。
免责声明：本文内容仅供参考，不代表本站观点。转载或引用请注明出处。
