使用python opencv對目錄下圖片進行去重的方法

更新時間：2019年01月12日 16:05:30 作者：Sand_Ng

今天小編就為大家分享一篇使用python opencv對目錄下圖片進行去重的方法，具有很好的參考價值，希望對大家有所幫助。一起跟隨小編過來看看吧

版本：

平臺：ubuntu 14 / I5 / 4G內(nèi)存

python版本：python2.7

opencv版本：2.13.4

依賴：

如果系統(tǒng)沒有python，則需要進行安裝

sudo apt-get install python

sudo apt-get install python-dev

sudo apt-get install python-pip

sudo pip install numpy mathplotlib

sudo apt-get install libcv-dev

sudo apt-get install python-opencv

使用感知哈希算法進行圖片去重

原理：對每個文件進行遍歷所有進行去重，因此圖片越多速度越慢，但是可以節(jié)省手動操作

感知哈希原理：

1、需要比較的圖片都縮放成8*8大小的灰度圖

2、獲得每個圖片每個像素與平均值的比較，得到指紋

3、根據(jù)指紋計算漢明距離

5、如果得出的不同的元素小于5則為相同（相似？）的圖片

#!/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)

為了節(jié)省操作，遍歷所有目錄，把想要去重的目錄遍歷一遍

#!/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

以上這篇使用python opencv對目錄下圖片進行去重的方法就是小編分享給大家的全部內(nèi)容了，希望能給大家一個參考，也希望大家多多支持腳本之家。

您可能感興趣的文章: