局部敏感哈希算法的代码

# -*- coding:utf-8 -*- import sys import math import time import string from opencv.cv import * from opencv.highgui import * Hash_1 = [1, 3, 7, 8] Hash_2 = [2, 6, 9, 14] Hash_3 = [3, 5, 15, 21] Hash_4 = [12, 17, 19, 23] #对SIFT特征向量进行归一化处理 def Make_Norm(array): total = 0 normalization = [0] * len(array) for i in range(len(array)): total += array[i] ** 2 total = total ** 0.5 if total != 0: for i in range(len(array)): normalization[i] = float(array[i]) / total return normalization #得到需要的12维颜色直方图向量 def Get_vector(image): width = image.width height = image.height vector = [0] * 12 index = 0 for m in range(2): for n in range(2): total = 0 for i in range(m * width / 2, (m+1) * width / 2): for j in range(n * height / 2, (n+1) * height / 2): for k in range(3): vector[index+k] += image[j,i][k] for k in range(3): total += vector[index+k] for k in range(3): vector[index+k] = vector[index+k] / total index += 3 return vector #对12维向量做量化处理 def Make_character(vector): for i in range(len(vector)): if vector[i] < 0.3: vector[i] = 0 elif vector[i] > 0.6: vector[i] = 2 else: vector[i] = 1 return vector #对向量在相应hash函数上求投影 def Make_projection(I, vector): section = {} projection = "" I.sort() for i in range(len(I)): tmp = (I[i] + 1) / 2 try: section[tmp].append(I[i]) except: section[tmp] = [] section[tmp].append(I[i]) for i in range(1, len(vector)+1): num = 0 try: current = section[i] for j in range(len(current)): if current[j] - 2 * (i - 1) <= vector[i-1]: num += 1 projection += (num * "1" + (len(current) - num) * "0") except: pass return projection #求两个图片对应的特征向量的相似度 def Get_similarity(point1, point2): totalSum = 0 for i in range(len(point1)): totalSum += point1[i] * point2[i] return totalSum #求每个向量投影后得到的hash值 def Hash_key(image): p = Get_vector(image) p = Make_character(p) tmp = [] tmp.append(Make_projection(Hash_1, p)) tmp.append(Make_projection(Hash_2, p)) tmp.append(Make_projection(Hash_3, p)) tmp.append(Make_projection(Hash_4, p)) return tmp def main(): result = [] image = cvLoadImage("target.jpg", 1) Object = Hash_key(image) p = Get_vector(image) start = time.clock() for i in range(1, 41): image = cvLoadImage(("%d.jpg" % i), 1) tmp = Hash_key(image) if tmp == Object: result.append(i) end = time.clock() print end - start start = time.clock() if len(result) >= 1: max = 0 for i in range(len(result)): image = cvLoadImage(("%d.jpg" % result[i]), 1) q = Get_vector(image) p = Make_Norm(p) q = Make_Norm(q) similarity = Get_similarity(p, q) if similarity > max: max = similarity picture = result[i] print "最好的匹配是：",("%d.jpg" % picture) else: print "最好的匹配是：",("%d.jpg" % result[0]) end = time.clock() print end - start if __name__ == '__main__': main()