因?yàn)檩斎胧且曨l，切完幀之后都是連續(xù)圖片，所以我的目錄結(jié)構(gòu)如下：

其中frame_output是視頻切幀后的保存路徑，1和2文件夾分別對(duì)應(yīng)兩個(gè)是視頻切幀后的圖片。

#encoding:utf-8import osimport sysimport cv2video_path = ’/home/pythonfile/video/’ # 絕對(duì)路徑，video下有兩段視頻out_frame_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), ’frame_output’) #frame_output是視頻切幀后的保存路徑if not os.path.exists(out_frame_path): os.makedirs(out_frame_path)print(’out_frame_path’, out_frame_path)files = []list1 = os.listdir(video_path)print(’list’, list1)for i in range(len(list1)): item = os.path.join(video_path, list1[i]) files.append(item)print(’files’,files)for k,file in enumerate(files): frame_dir = os.path.join(out_frame_path, ’%d’%(k+1)) if not os.path.exists(frame_dir):os.makedirs(frame_dir) cap = cv2.VideoCapture(file) j = 0 print(’start prossing NO.%d video’ % (k + 1)) while True:ret, frame = cap.read()j += 1if ret:#每三幀保存一張 if j % 3 == 0:cv2.imwrite(os.path.join(frame_dir, ’%d.jpg’%j), frame)else: cap.release() break print(’prossed NO.%d video’%(k+1))3. 刪除相似度高的圖片

# coding: utf-8import osimport cv2# from skimage.measure import compare_ssim# from skimage.metrics import _structural_similarityfrom skimage.metrics import structural_similarity as ssimdef delete(filename1): os.remove(filename1)def list_all_files(root): files = [] list = os.listdir(root) # os.listdir()方法：返回指定文件夾包含的文件或子文件夾名字的列表。該列表順序以字母排序 for i in range(len(list)):element = os.path.join(root, list[i])# 需要先使用python路徑拼接os.path.join()函數(shù)，將os.listdir()返回的名稱拼接成文件或目錄的絕對(duì)路徑再傳入os.path.isdir()和os.path.isfile().if os.path.isdir(element): # os.path.isdir()用于判斷某一對(duì)象(需提供絕對(duì)路徑)是否為目錄 # temp_dir = os.path.split(element)[-1] # os.path.split分割文件名與路徑,分割為data_dir和此路徑下的文件名，[-1]表示只取data_dir下的文件名 files.append(list_all_files(element))elif os.path.isfile(element): files.append(element) # print(’2’,files) return filesdef ssim_compare(img_files): count = 0 for currIndex, filename in enumerate(img_files):if not os.path.exists(img_files[currIndex]): print(’not exist’, img_files[currIndex]) breakimg = cv2.imread(img_files[currIndex])img1 = cv2.imread(img_files[currIndex + 1])#進(jìn)行結(jié)構(gòu)性相似度判斷# ssim_value = _structural_similarity.structural_similarity(img,img1,multichannel=True)ssim_value = ssim(img,img1,multichannel=True)if ssim_value > 0.9: #基數(shù) count += 1 imgs_n.append(img_files[currIndex + 1]) print(’big_ssim:’,img_files[currIndex], img_files[currIndex + 1], ssim_value)# 避免數(shù)組越界if currIndex+1 >= len(img_files)-1: break return countif __name__ == ’__main__’: path = ’/home/dj/pythonfile/frame_output/’ img_path = path imgs_n = [] all_files = list_all_files(path) #返回包含完整路徑的所有圖片名的列表 print(’1’,len(all_files)) for files in all_files:# 根據(jù)文件名排序，x.rfind(’/’)是從右邊尋找第一個(gè)‘/’出現(xiàn)的位置，也就是最后出現(xiàn)的位置# 注意sort和sorted的區(qū)別，sort作用于原列表，sorted生成新的列表，且sorted可以作用于所有可迭代對(duì)象files.sort(key = lambda x: int(x[x.rfind(’/’)+1:-4]))#路徑中包含“/”# print(files)img_files = []for img in files: if img.endswith(’.jpg’):# 將所有圖片名都放入列表中img_files.append(img)count = ssim_compare(img_files)print(img[:img.rfind(’/’)],'路徑下刪除的圖片數(shù)量為：',count) for image in imgs_n:delete(image)4. 導(dǎo)入skimage.measure import compare_ssim出錯(cuò)的解決方法：

將

from skimage.measure import compare_ssim

改為

from skimage.metrics import _structural_similarity5. structural_similarity.py的源碼

from warnings import warnimport numpy as npfrom scipy.ndimage import uniform_filter, gaussian_filterfrom ..util.dtype import dtype_rangefrom ..util.arraycrop import cropfrom .._shared.utils import warn, check_shape_equality__all__ = [’structural_similarity’]def structural_similarity(im1, im2, *, win_size=None, gradient=False, data_range=None, multichannel=False, gaussian_weights=False, full=False, **kwargs): ''' Compute the mean structural similarity index between two images. Parameters ---------- im1, im2 : ndarrayImages. Any dimensionality with same shape. win_size : int or None, optionalThe side-length of the sliding window used in comparison. Must be anodd value. If `gaussian_weights` is True, this is ignored and thewindow size will depend on `sigma`. gradient : bool, optionalIf True, also return the gradient with respect to im2. data_range : float, optionalThe data range of the input image (distance between minimum andmaximum possible values). By default, this is estimated from the imagedata-type. multichannel : bool, optionalIf True, treat the last dimension of the array as channels. Similaritycalculations are done independently for each channel then averaged. gaussian_weights : bool, optionalIf True, each patch has its mean and variance spatially weighted by anormalized Gaussian kernel of width sigma=1.5. full : bool, optionalIf True, also return the full structural similarity image. Other Parameters ---------------- use_sample_covariance : boolIf True, normalize covariances by N-1 rather than, N where N is thenumber of pixels within the sliding window. K1 : floatAlgorithm parameter, K1 (small constant, see [1]_). K2 : floatAlgorithm parameter, K2 (small constant, see [1]_). sigma : floatStandard deviation for the Gaussian when `gaussian_weights` is True. Returns ------- mssim : floatThe mean structural similarity index over the image. grad : ndarrayThe gradient of the structural similarity between im1 and im2 [2]_.This is only returned if `gradient` is set to True. S : ndarrayThe full SSIM image. This is only returned if `full` is set to True. Notes ----- To match the implementation of Wang et. al. [1]_, set `gaussian_weights` to True, `sigma` to 1.5, and `use_sample_covariance` to False. .. versionchanged:: 0.16This function was renamed from ``skimage.measure.compare_ssim`` to``skimage.metrics.structural_similarity``. References ---------- .. [1] Wang, Z., Bovik, A. C., Sheikh, H. R., & Simoncelli, E. P. (2004). Image quality assessment: From error visibility to structural similarity. IEEE Transactions on Image Processing, 13, 600-612. https://ece.uwaterloo.ca/~z70wang/publications/ssim.pdf, :DOI:`10.1109/TIP.2003.819861` .. [2] Avanaki, A. N. (2009). Exact global histogram specification optimized for structural similarity. Optical Review, 16, 613-621. :arxiv:`0901.0065` :DOI:`10.1007/s10043-009-0119-z` ''' check_shape_equality(im1, im2) if multichannel:# loop over channelsargs = dict(win_size=win_size, gradient=gradient, data_range=data_range, multichannel=False, gaussian_weights=gaussian_weights, full=full)args.update(kwargs)nch = im1.shape[-1]mssim = np.empty(nch)if gradient: G = np.empty(im1.shape)if full: S = np.empty(im1.shape)for ch in range(nch): ch_result = structural_similarity(im1[..., ch], im2[..., ch], **args) if gradient and full:mssim[..., ch], G[..., ch], S[..., ch] = ch_result elif gradient:mssim[..., ch], G[..., ch] = ch_result elif full:mssim[..., ch], S[..., ch] = ch_result else:mssim[..., ch] = ch_resultmssim = mssim.mean()if gradient and full: return mssim, G, Selif gradient: return mssim, Gelif full: return mssim, Selse: return mssim K1 = kwargs.pop(’K1’, 0.01) K2 = kwargs.pop(’K2’, 0.03) sigma = kwargs.pop(’sigma’, 1.5) if K1 < 0:raise ValueError('K1 must be positive') if K2 < 0:raise ValueError('K2 must be positive') if sigma < 0:raise ValueError('sigma must be positive') use_sample_covariance = kwargs.pop(’use_sample_covariance’, True) if gaussian_weights:# Set to give an 11-tap filter with the default sigma of 1.5 to match# Wang et. al. 2004.truncate = 3.5 if win_size is None:if gaussian_weights: # set win_size used by crop to match the filter size r = int(truncate * sigma + 0.5) # radius as in ndimage win_size = 2 * r + 1else: win_size = 7 # backwards compatibility if np.any((np.asarray(im1.shape) - win_size) < 0):raise ValueError( 'win_size exceeds image extent. If the input is a multichannel ' '(color) image, set multichannel=True.') if not (win_size % 2 == 1):raise ValueError(’Window size must be odd.’) if data_range is None:if im1.dtype != im2.dtype: warn('Inputs have mismatched dtype. Setting data_range based on ' 'im1.dtype.', stacklevel=2)dmin, dmax = dtype_range[im1.dtype.type]data_range = dmax - dmin ndim = im1.ndim if gaussian_weights:filter_func = gaussian_filterfilter_args = {’sigma’: sigma, ’truncate’: truncate} else:filter_func = uniform_filterfilter_args = {’size’: win_size} # ndimage filters need floating point data im1 = im1.astype(np.float64) im2 = im2.astype(np.float64) NP = win_size ** ndim # filter has already normalized by NP if use_sample_covariance:cov_norm = NP / (NP - 1) # sample covariance else:cov_norm = 1.0 # population covariance to match Wang et. al. 2004 # compute (weighted) means ux = filter_func(im1, **filter_args) uy = filter_func(im2, **filter_args) # compute (weighted) variances and covariances uxx = filter_func(im1 * im1, **filter_args) uyy = filter_func(im2 * im2, **filter_args) uxy = filter_func(im1 * im2, **filter_args) vx = cov_norm * (uxx - ux * ux) vy = cov_norm * (uyy - uy * uy) vxy = cov_norm * (uxy - ux * uy) R = data_range C1 = (K1 * R) ** 2 C2 = (K2 * R) ** 2 A1, A2, B1, B2 = ((2 * ux * uy + C1, 2 * vxy + C2, ux ** 2 + uy ** 2 + C1, vx + vy + C2)) D = B1 * B2 S = (A1 * A2) / D # to avoid edge effects will ignore filter radius strip around edges pad = (win_size - 1) // 2 # compute (weighted) mean of ssim mssim = crop(S, pad).mean() if gradient:# The following is Eqs. 7-8 of Avanaki 2009.grad = filter_func(A1 / D, **filter_args) * im1grad += filter_func(-S / B2, **filter_args) * im2grad += filter_func((ux * (A2 - A1) - uy * (B2 - B1) * S) / D, **filter_args)grad *= (2 / im1.size)if full: return mssim, grad, Selse: return mssim, grad else:if full: return mssim, Selse: return mssim

到此這篇關(guān)于教你怎么用python刪除相似度高的圖片的文章就介紹到這了,更多相關(guān)python刪除相似度高的圖片內(nèi)容請(qǐng)搜索好吧啦網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章希望大家以后多多支持好吧啦網(wǎng)！