计算两幅图像的相似度（PSNR、SSIM、MSE、余弦相似度、MD5、直方图、互信息、Hash）& 代码实现 与举例 |
您所在的位置：网站首页 › 余弦相似度计算文本相似度怎么算 › 计算两幅图像的相似度（PSNR、SSIM、MSE、余弦相似度、MD5、直方图、互信息、Hash）& 代码实现 与举例 |
计算两幅图像的相似度（PSNR、SSIM、MSE、余弦相似度、MD5、直方图、互信息、Hash）& 代码实现 与举例
|
MSE（Mean Squared Error）均方误差
MSE公式
MSE 计算模型的预测 Ŷ 与真实标签 Y 的接近程度。公式表示为：
对于两个m×n的单通道图像I和K，它们的均方误差可定义为：
优点：MSE的函数曲线光滑、连续，处处可导，便于使用梯度下降算法，是一种常用的损失函数。而且，随着误差的减小，梯度也在减小，这有利于收敛，即使使用固定的学习速率，也能较快的收敛到最小值。 缺点：当真实值y和预测值f(x)的差值大于1时，会放大误差；而当差值小于1时，则会缩小误差，这是平方运算决定的。MSE对于较大的误差(>1)给予较大的惩罚，较小的误差( avreage: hash.append(1) else: hash.append(0) return hash #均值哈希算法 def aHash(image): #缩放为8*8 image=cv2.resize(image,(8,8),interpolation=cv2.INTER_CUBIC) #转换为灰度图 image=cv2.cvtColor(image,cv2.COLOR_BGR2GRAY) avreage = np.mean(image) hash = [] for i in range(image.shape[0]): for j in range(image.shape[1]): if image[i,j] > avreage: hash.append(1) else: hash.append(0) return hash #差值感知算法 def dHash(image): #缩放9*8 image=cv2.resize(image,(9,8),interpolation=cv2.INTER_CUBIC) #转换灰度图 image=cv2.cvtColor(image,cv2.COLOR_BGR2GRAY) # print(image.shape) hash=[] #每行前一个像素大于后一个像素为1，相反为0，生成哈希 for i in range(8): for j in range(8): if image[i,j]>image[i,j+1]: hash.append(1) else: hash.append(0) return hash #计算汉明距离 def Hamming_distance(hash1,hash2): num = 0 for index in range(len(hash1)): if hash1[index] != hash2[index]: num += 1 return num if __name__ == "__main__": image_file1 = './data/cartoon1.jpg' image_file2 = './data/cartoon3.jpg' img1 = cv2.imread(image_file1) img2 = cv2.imread(image_file2) hash1 = pHash(img1) hash2 = pHash(img2) dist = Hamming_distance(hash1, hash2) #将距离转化为相似度 similarity = 1 - dist * 1.0 / 64 print(dist) print(similarity) 直方图距离图片的相似度--直方图距离 - 知乎 (zhihu.com) 直方图距离描述方法描述：按照某种距离度量的标准对两幅图像的直方图进行相似度的测量。 图像直方图丰富的图像细节信息，反映了图像像素点的概率分布情况，统计每一个像素点强度值具有的像素个数。 优点：计算量比较小。缺点： 直方图反应的是图像灰度值得概率分布，并没有图像的空间位置信息在里面，因此，会出现误判；比如纹理结构相同，但明暗不同的图像，应该相似度很高，但实际结果是相似度很低，而纹理结构不同，但明暗相近的图像，相似度却很高。计算步骤： 将图片resize，得到相同大小的图片；将图片灰度，灰度后图片的像素在[0-255]之间；计算图片的直方图数据，统计相同像素点的概率分布；根据相关性计算公式，计算两个图片直方图的相关性。 代码实现 import cv2 def calculate(image1, image2): # 灰度直方图算法 # 计算单通道的直方图的相似值 hist1 = cv2.calcHist([image1], [0], None, [256], [0.0, 255.0]) hist2 = cv2.calcHist([image2], [0], None, [256], [0.0, 255.0]) # 计算直方图的重合度 degree = 0 for i in range(len(hist1)): if hist1[i] != hist2[i]: degree = degree + (1 - abs(hist1[i] - hist2[i]) / max(hist1[i], hist2[i])) else: degree = degree + 1 degree = degree / len(hist1) return degree def classify_hist_with_split(image1, image2, size=(256, 256)): # RGB每个通道的直方图相似度 # 将图像resize后，分离为RGB三个通道，再计算每个通道的相似值 image1 = cv2.resize(image1, size) image2 = cv2.resize(image2, size) sub_image1 = cv2.split(image1) sub_image2 = cv2.split(image2) sub_data = 0 for im1, im2 in zip(sub_image1, sub_image2): sub_data += calculate(im1, im2) sub_data = sub_data / 3 return sub_data 互信息（Mutual Information）归一化互信息(NMI)评价指标_易_的博客-CSDN博客_nmi指标 信息熵、相对熵、互信息信息熵：对信息进行量化度量。可以理解为某种特定信息的出现概率。
相对熵：（relative entropy），又被称为Kullback-Leibler散度（Kullback-Leibler divergence，KL散度）或信息散度（information divergence），是两个概率分布（probability distribution）间差异的非对称性度量 。在在信息理论中，相对熵等价于两个概率分布的信息熵（Shannon entropy）的差值。 互信息：(Mutual Information)是信息论里一种有用的信息度量，它可以看成是一个随机变量中包含的关于另一个随机变量的信息量，或者说是一个随机变量由于已知另一个随机变量而减少的不肯定性。
归一化互信息：将互信息放在[0,1]之间，容易评价算法的好坏。
GitHub - whtsky/pixelmatch-py: A fast pixel-level image comparison library, originally created to compare screenshots in tests. 利用像素之间的匹配来计算相似度 # https://github.com/whtsky/pixelmatch-py # 第一步： pip install pixelmatch # 第二步： from PIL import Image from pixelmatch.contrib.PIL import pixelmatch img_a = Image.open("a.png") img_b = Image.open("b.png") img_diff = Image.new("RGBA", img_a.size) # note how there is no need to specify dimensions mismatch = pixelmatch(img_a, img_b, img_diff, includeAA=True) img_diff.save("diff.png") 举例1——使用四种方法计算图片相似度：MD5、直方图、PSNR、SSIM有史以来最全的图像相似度算法_小殊小殊的博客-CSDN博客_图片相似度 使用四种方法计算图片相似度：MD5（粗暴的md5比较，判断是否完全相同）、直方图、PSNR、SSIM。 import numpy as np from PIL import Image from skimage.metrics import structural_similarity import cv2 import os import hashlib import math ''' 粗暴的md5比较 返回是否完全相同 ''' def md5_similarity(img1_path, img2_path): file1 = open(img1_path, "rb") file2 = open(img2_path, "rb") md = hashlib.md5() md.update(file1.read()) res1 = md.hexdigest() md = hashlib.md5() md.update(file2.read()) res2 = md.hexdigest() return res1 == res2 def normalize(data): return data / np.sum(data) ''' 直方图相似度 相关性比较 cv2.HISTCMP_CORREL：值越大，相似度越高 相交性比较 cv2.HISTCMP_INTERSECT：值越大，相似度越高 卡方比较 cv2.HISTCMP_CHISQR：值越小，相似度越高 巴氏距离比较 cv2.HISTCMP_BHATTACHARYYA：值越小，相似度越高 ''' def hist_similarity(img1, img2, hist_size=256): imghistb1 = cv2.calcHist([img1], [0], None, [hist_size], [0, 256]) imghistg1 = cv2.calcHist([img1], [1], None, [hist_size], [0, 256]) imghistr1 = cv2.calcHist([img1], [2], None, [hist_size], [0, 256]) imghistb2 = cv2.calcHist([img2], [0], None, [hist_size], [0, 256]) imghistg2 = cv2.calcHist([img2], [1], None, [hist_size], [0, 256]) imghistr2 = cv2.calcHist([img2], [2], None, [hist_size], [0, 256]) distanceb = cv2.compareHist(normalize(imghistb1), normalize(imghistb2), cv2.HISTCMP_CORREL) distanceg = cv2.compareHist(normalize(imghistg1), normalize(imghistg2), cv2.HISTCMP_CORREL) distancer = cv2.compareHist(normalize(imghistr1), normalize(imghistr2), cv2.HISTCMP_CORREL) meandistance = np.mean([distanceb, distanceg, distancer]) return meandistance def PSNR(img1, img2): mse = np.mean((img1/255. - img2/255.) ** 2) if mse == 0: return 100 PIXEL_MAX = 1 return 20 * math.log10(PIXEL_MAX / math.sqrt(mse)) def SSIM(img1, img2): gray1 = cv2.cvtColor(img1, cv2.COLOR_BGR2GRAY) gray2 = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY) # 计算两个灰度图像之间的结构相似度 score, diff = structural_similarity(gray1, gray2, win_size=101, full=True) # diff = (diff * 255).astype("uint8") # print("SSIM:{}".format(score)) return score, diff def ssim(img1, img2): u_true = np.mean(img1) u_pred = np.mean(img2) var_true = np.var(img1) var_pred = np.var(img2) std_true = np.sqrt(var_true) std_pred = np.sqrt(var_pred) c1 = np.square(0.01*7) c2 = np.square(0.03*7) ssim = (2 * u_true * u_pred + c1) * (2 * std_pred * std_true + c2) denom = (u_true ** 2 + u_pred ** 2 + c1) * (var_pred + var_true + c2) return ssim/denom def MSE(img1,img2): mse = np.mean( (img1 - img2) ** 2 ) return mse if __name__ == '__main__': img1_path = '1_08.jpg' img2_path = '2_08.jpg' img1 = cv2.imread(img1_path) img2 = cv2.imread(img2_path) # 1.粗暴的md5比较 返回是否完全相同 print('md5_similarity:', md5_similarity(img1_path, img2_path)) # 2.直方图相似度 print('hist_similarity:', hist_similarity(img1, img2)) # 3.PSNR print('PSNR:', PSNR(img1, img2)) # 4.SSIM print('灰度图SSIM:', SSIM(img1, img2)[0]) print('RGB图_ssim:', ssim(img1, img2)) print('RGB图_structural_similarity:', structural_similarity(img1,img2, multichannel=True)) # 5.MSE print('MSE:', MSE(img1, img2))效果展示： 
两图
md5_similarity: False
hist_similarity: 0.1344447074377748
PSNR: 12.980832701995697
灰度图SSIM: 0.5966066785284522
RGB图_ssim: 0.9917382398395064
RGB图_structural_similarity: 0.525279441002218
MSE: 86.05895699395074
举例2——批量计算PSNR、SSIM、MSE
import os
import numpy as np
from glob import glob
import cv2
from skimage.measure import compare_mse,compare_ssim,compare_psnr
def read_img(path):
return cv2.imread(path,cv2.IMREAD_GRAYSCALE)
def mse(tf_img1, tf_img2):
return compare_mse(tf_img1,tf_img2)
def psnr(tf_img1, tf_img2):
return compare_psnr(tf_img1,tf_img2)
def ssim(tf_img1, tf_img2):
return compare_ssim(tf_img1,tf_img2)
def main():
WSI_MASK_PATH1 = 'E:/test/A/'
WSI_MASK_PATH2 = 'E:/test/B/'
path_real = glob(os.path.join(WSI_MASK_PATH1, '*.jpg'))
path_fake = glob(os.path.join(WSI_MASK_PATH2, '*.jpg'))
list_psnr = []
list_ssim = []
list_mse = []
for i in range(len(path_real)):
t1 = read_img(path_real[i])
t2 = read_img(path_fake[i])
result1 = np.zeros(t1.shape,dtype=np.float32)
result2 = np.zeros(t2.shape,dtype=np.float32)
cv2.normalize(t1,result1,alpha=0,beta=1,norm_type=cv2.NORM_MINMAX,dtype=cv2.CV_32F)
cv2.normalize(t2,result2,alpha=0,beta=1,norm_type=cv2.NORM_MINMAX,dtype=cv2.CV_32F)
mse_num = mse(result1, result2)
psnr_num = psnr(result1, result2)
ssim_num = ssim(result1, result2)
list_psnr.append(psnr_num)
list_ssim.append(ssim_num)
list_mse.append(mse_num)
#输出每张图像的指标：
print("{}/".format(i+1)+"{}:".format(len(path_real)))
str = "\\"
print("image:"+path_real[i][(path_real[i].index(str)+1):])
print("PSNR:", psnr_num)
print("SSIM:", ssim_num)
print("MSE:",mse_num)
#输出平均指标：
print("平均PSNR:", np.mean(list_psnr)) # ,list_psnr)
print("平均SSIM:", np.mean(list_ssim)) # ,list_ssim)
print("平均MSE:", np.mean(list_mse)) # ,list_mse)
if __name__ == '__main__':
main()
举例3——计算Y通道或RGB通道的PSNR和SSIM
Python计算Y通道或者RGB通道的PSNR_SSIM_未知量0520的博客-CSDN博客_folder_gt 添加：PSNR/SSIM计算结果，保存文件（txt），便于后期导入excel列表分析。代码介绍：PSNR_SSIM效果和matlab计算的结果一致（本人没有验证）。代码功能，能够计算图像的Y_channel 或者RGB_channel状态下的PSNR_SSIM结果用法：将真实图片，生成图片分别放入两个文件夹，代码中有选择两种方式（only_Y, RGB），切换注释相应的代码行即可。注意：安装相应的代码库。 ''' calculate the PSNR and SSIM. same as MATLAB's results https://blog.csdn.net/qq_45041702/article/details/120997743 ''' import os import math import numpy as np import cv2 import glob import os def main(): # Configurations # GT - Ground-truth; # Gen: Generated / Restored / Recovered images folder_GT = '/content/drive/MyDrive/Experiment/codes/26_PSNR_SSIM/All/groud_truth_png' folder_Gen = '/content/drive/MyDrive/Experiment/codes/26_PSNR_SSIM/All/1_ESPCN_png' #save the psnr and ssim score by txt PS_path = '/content/drive/MyDrive/Experiment/codes/26_PSNR_SSIM/Score' if not os.path.exists(PS_path): print('NO_path to sava the NIQE_Score,Making....') os.makedirs(PS_path) else: print("The NIQE_Score_path has existed") PS_txt = open(PS_path +'/score1.txt', 'a' ) crop_border = 4 # same with scale suffix = '' # suffix for Gen images test_Y = False # True: test Y channel only; False: test RGB channels PSNR_all = [] SSIM_all = [] img_list = sorted(glob.glob(folder_GT + '/*')) if test_Y: print('Testing Y channel.') else: print('Testing RGB channels.') for i, img_path in enumerate(img_list): base_name = os.path.splitext(os.path.basename(img_path))[0] print(base_name) im_GT = cv2.imread(img_path) / 255. #不同格式图像 # im_Gen = cv2.imread(os.path.join(folder_Gen, base_name + suffix + '.tif')) / 255. im_Gen = cv2.imread(os.path.join(folder_Gen, base_name + suffix + '.png')) / 255. if test_Y and im_GT.shape[2] == 3: # evaluate on Y channel in YCbCr color space im_GT_in = bgr2ycbcr(im_GT) im_Gen_in = bgr2ycbcr(im_Gen) else: im_GT_in = im_GT im_Gen_in = im_Gen # crop borders if crop_border == 0: cropped_GT = im_GT_in cropped_Gen = im_Gen_in else: if im_GT_in.ndim == 3: cropped_GT = im_GT_in[crop_border:-crop_border, crop_border:-crop_border, :] cropped_Gen = im_Gen_in[crop_border:-crop_border, crop_border:-crop_border, :] elif im_GT_in.ndim == 2: cropped_GT = im_GT_in[crop_border:-crop_border, crop_border:-crop_border] cropped_Gen = im_Gen_in[crop_border:-crop_border, crop_border:-crop_border] else: raise ValueError('Wrong image dimension: {}. Should be 2 or 3.'.format(im_GT_in.ndim)) #不同通道数（Y通道和RGB三个通道），需要更改 # calculate PSNR and SSIM # PSNR = calculate_psnr(cropped_GT * 255, cropped_Gen * 255) PSNR = calculate_rgb_psnr(cropped_GT * 255, cropped_Gen * 255) SSIM = calculate_ssim(cropped_GT * 255, cropped_Gen * 255) print('{:3d} - {:25}. \tPSNR: {:.4f} dB, \tSSIM: {:.4f}'.format( i + 1, base_name, PSNR, SSIM)) PSNR_all.append(PSNR) SSIM_all.append(SSIM) single_info = '[{}],PSNR(dB),{:.4f}, SSIM,{:.4f}' PS_txt.write(single_info.format(base_name,PSNR,SSIM)) PS_txt.write("\n") Mean_format = 'Mean_PSNR: {:.4f}, Mean_SSIM: {:.4f}' Mean_PSNR = sum(PSNR_all) / len(PSNR_all) Mean_SSIM = sum(SSIM_all) / len(SSIM_all) print(Mean_format.format(Mean_PSNR,Mean_SSIM)) PS_txt.write(Mean_format.format(Mean_PSNR,Mean_SSIM)) PS_txt.write('\n') def calculate_psnr(img1, img2): # img1 and img2 have range [0, 255] img1 = img1.astype(np.float64) img2 = img2.astype(np.float64) mse = np.mean((img1 - img2)**2) if mse == 0: return float('inf') return 20 * math.log10(255.0 / math.sqrt(mse)) def calculate_rgb_psnr(img1, img2): """calculate psnr among rgb channel, img1 and img2 have range [0, 255] """ n_channels = np.ndim(img1) sum_psnr = 0 for i in range(n_channels): this_psnr = calculate_psnr(img1[:,:,i], img2[:,:,i]) sum_psnr += this_psnr return sum_psnr/n_channels def ssim(img1, img2): C1 = (0.01 * 255)**2 C2 = (0.03 * 255)**2 img1 = img1.astype(np.float64) img2 = img2.astype(np.float64) kernel = cv2.getGaussianKernel(11, 1.5) window = np.outer(kernel, kernel.transpose()) mu1 = cv2.filter2D(img1, -1, window)[5:-5, 5:-5] # valid mu2 = cv2.filter2D(img2, -1, window)[5:-5, 5:-5] mu1_sq = mu1**2 mu2_sq = mu2**2 mu1_mu2 = mu1 * mu2 sigma1_sq = cv2.filter2D(img1**2, -1, window)[5:-5, 5:-5] - mu1_sq sigma2_sq = cv2.filter2D(img2**2, -1, window)[5:-5, 5:-5] - mu2_sq sigma12 = cv2.filter2D(img1 * img2, -1, window)[5:-5, 5:-5] - mu1_mu2 ssim_map = ((2 * mu1_mu2 + C1) * (2 * sigma12 + C2)) / ((mu1_sq + mu2_sq + C1) * (sigma1_sq + sigma2_sq + C2)) return ssim_map.mean() def calculate_ssim(img1, img2): '''calculate SSIM the same outputs as MATLAB's img1, img2: [0, 255] ''' if not img1.shape == img2.shape: raise ValueError('Input images must have the same dimensions.') if img1.ndim == 2: return ssim(img1, img2) elif img1.ndim == 3: if img1.shape[2] == 3: ssims = [] for i in range(img1.shape[2]): ssims.append(ssim(img1[..., i], img2[..., i])) return np.array(ssims).mean() elif img1.shape[2] == 1: return ssim(np.squeeze(img1), np.squeeze(img2)) else: raise ValueError('Wrong input image dimensions.') def bgr2ycbcr(img, only_y=True): '''same as matlab rgb2ycbcr only_y: only return Y channel Input: uint8, [0, 255] float, [0, 1] ''' in_img_type = img.dtype img.astype(np.float32) if in_img_type != np.uint8: img *= 255. # convert if only_y: rlt = np.dot(img, [24.966, 128.553, 65.481]) / 255.0 + 16.0 else: rlt = np.matmul(img, [[24.966, 112.0, -18.214], [128.553, -74.203, -93.786], [65.481, -37.797, 112.0]]) / 255.0 + [16, 128, 128] if in_img_type == np.uint8: rlt = rlt.round() else: rlt /= 255. return rlt.astype(in_img_type) if __name__ == '__main__': main()结果展示：
SSIM---结构相似性算法（OpenCV+Python)_hedgehog__的博客-CSDN博客_python 结构相似性 通过调用skimage.metrics包下的SSIM算法，结合OpenCV的阈值分割及轮廓提取算法，找出两幅图像的差异。 # https://blog.csdn.net/hedgehog__/article/details/107257755 import cv2 import imutils from skimage.metrics import structural_similarity import time from skimage import filters, img_as_ubyte import numpy as np start = time.time() # 读入图像，转为灰度图像 src = cv2.imread('C:/Users/Hedgehog/Desktop/right.jpg') img = cv2.imread('C:/Users/Hedgehog/Desktop/left.jpg') grayA = cv2.cvtColor(src, cv2.COLOR_BGR2GRAY) grayB = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 计算两个灰度图像之间的结构相似度 (score, diff) = structural_similarity(grayA, grayB, win_size=101, full=True) diff = (diff * 255).astype("uint8") cv2.namedWindow("diff", cv2.WINDOW_NORMAL) cv2.imshow("diff", diff) print("SSIM:{}".format(score)) # 找到不同的轮廓以致于可以在表示为 '不同'的区域放置矩形 # 全局自适应阈值分割（二值化），返回值有两个，第一个是阈值，第二个是二值图像 dst = cv2.threshold(diff, 0, 255, cv2.THRESH_BINARY_INV | cv2.THRESH_OTSU)[1] cv2.namedWindow("threshold", cv2.WINDOW_NORMAL) cv2.imshow('threshold', dst) # findContours找轮廓，返回值有两个，第一个是轮廓信息，第二个是轮廓的层次信息（“树”状拓扑结构） # cv2.RETR_EXTERNAL：只检测最外层轮廓 # cv2.CHAIN_APPROX_SIMPLE：压缩水平方向、垂直方向和对角线方向的元素，保留该方向的终点坐标，如矩形的轮廓可用4个角点表示 contours, hierarchy = cv2.findContours(dst.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE) print(contours) newimg = np.zeros(dst.shape, np.uint8) # 定义一个和图像分割处理后相同大小的黑色图 # drawContours画轮廓，将找到的轮廓信息画出来 cv2.drawContours(newimg, contours, -1, (255, 255, 255), 1) cv2.namedWindow("contours", cv2.WINDOW_NORMAL) cv2.imshow('contours', newimg) # cnts = cnts[0] if imutils.is_cv3() else cnts[0] 取findContours函数的第一个返回值，即取轮廓信息 # 找到一系列区域，在区域周围放置矩形 for c in contours: (x, y, w, h) = cv2.boundingRect(c) # boundingRect函数：计算轮廓的垂直边界最小矩形，矩形是与图像上下边界平行的 cv2.rectangle(src, (x, y), (x + w, y + h), (0, 255, 0), 2) # rectangle函数：使用对角线的两点pt1，pt2画一个矩形轮廓 cv2.rectangle(img, (x, y), (x + w, y + h), (0, 255, 0), 2) # 画矩形的图, pt1, pt2,(对角线两点的坐标), 矩形边框的颜色，矩形边框的粗细 end = time.time() print(end - start) # 用cv2.imshow 展现最终对比之后的图片 cv2.namedWindow("right", cv2.WINDOW_NORMAL) cv2.imshow('right', src) cv2.namedWindow("left", cv2.WINDOW_NORMAL) cv2.imshow('left', img) cv2.waitKey(0) cv2.destroyAllWindows()效果展示：
计算两幅图像的相似度总结_JerrySing的博客-CSDN博客_compare_ssim Python计算图片之间的相似度_～小疯子～的博客-CSDN博客_python 识别图片相似度 有史以来最全的图像相似度算法_小殊小殊的博客-CSDN博客_图片相似度 （超详细）实现计算图片相似度MSE和PSNR_Natuki丶的博客-CSDN博客_计算psnr和mse Python批量计算PSNR、SSIM、MSE_杨杨杨Garrick的博客-CSDN博客 图像质量评价指标MSE/PSNR/SSIM_秋天的从从的博客-CSDN博客_图像mse 图片相相似度计算（Hash、SSIM、compareHist）_南苏月的博客-CSDN博客_图像相似度计算公式 【OpenCV-Python】：图像PSNR、SSIM、MSE计算_米开朗琪罗~的博客-CSDN博客_图像的mse计算 Python计算Y通道或者RGB通道的PSNR_SSIM_未知量0520的博客-CSDN博客_folder_gt scikit-image 0.18.0版本计算PSNR、SSIM、MSE(Python代码)_YZBshanshan的博客-CSDN博客_scikit-image版本 图像质量评估中的PSNR和SSIM的定义，公式和含义_·如烟·的博客-CSDN博客_psnr定义 【基础知识】图像去噪评估—PSNR和SSIM - 梁君牧 - 博客园 (cnblogs.com) SSIM---结构相似性算法（OpenCV+Python)_hedgehog__的博客-CSDN博客_python 结构相似性 图片相似度计算方法总结 - 知乎 (zhihu.com) 相似度计算方法(三) 余弦相似度_潘永青的博客-CSDN博客_余弦相似度 图片的相似度--直方图距离 - 知乎 (zhihu.com) |
【本文地址】
今日新闻 |
推荐新闻 |