實現(xiàn)效果圖：

同時還使用了 mask圖，要識別的區(qū)域為白色，背景為黑色

import cv2 import numpy as np import os.path as osp import os import numpy as np from tqdm import tqdm from skimage import morphology from skimage import segmentationdef get_ori_list(ori_folder):img_list = os.listdir(ori_folder)ori_list = []for img_name in img_list:flag = 0 for sample in check_list:if sample in img_name:flag=1breakif flag==0:ori_list.append(osp.join(ori_folder,img_name))if len(ori_list)>20:breakreturn ori_listcheck_list = ['copper','bg','check','dust'] """超像素由一系列位置相鄰且顏色、亮度、紋理等特征相似的像素點組成的小區(qū)域。 這些小區(qū)域大多保留了進一步進行圖像分割的有效信息，且一般不會破壞圖像中物體的邊界信息， 用少量的超像素代替大量像素表達圖像特征，降低了圖像處理的復雜度， 一般作為分割算法的預處理步驟。"""def get_reverse(pic_matrix):where_0 = np.where(pic_matrix == 0)where_255 = np.where(pic_matrix == 255)pic_matrix[where_0] = 255pic_matrix[where_255] = 0return pic_matrixdef use_slic_by_SLIC(img_path,mask_path,end_path):"""https://scikit-image.org/docs/dev/auto_examples/segmentation/plot_mask_slic.html#sphx-glr-auto-examples-segmentation-plot-mask-slic-py"""image = cv2.imread(img_path)#image = cv2.resize(image,(256,256))mask_image = cv2.imread(mask_path)#mask_image = get_reverse(mask_image)# kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (2, 2))# mask_image = cv2.dilate(mask_image, kernel)#mask_image = cv2.resize(mask_image,(256,256))img_R = mask_image[:,:,0]mask = img_R>220# 生成扁平的盤狀結構元素 skimage.morphology.diskmask = morphology.opening(mask, morphology.disk(10))#mask是一個只包含True和False的ndarray，它的shape和data一致，# segmentation.slic在Color-（x，y，z）空間中使用k-means聚類來分割圖像。m_slic = segmentation.slic(image, n_segments=20000, mask=mask)slic_image = segmentation.mark_boundaries(image, m_slic,outline_color=(0,1,1))cv2.imwrite(end_path,slic_image*255) return slic_image*255if __name__ == '__main__':ori_folder = '/cloud_disk/users/huh/dataset/PCB/ddrnet_23_slim/pre_process_img'end_folder = '/cloud_disk/users/huh/pcb/script/slic_result/1_cv2'ori_list = get_ori_list(ori_folder)for img_path in tqdm(ori_list):end_path = osp.join(end_folder,osp.basename(img_path))mask_path = osp.join(ori_folder,osp.basename(img_path[:-4])+'_bg_mask.jpg')try:slic_image = use_slic_by_SLIC(img_path,mask_path,end_path)# slic_image = cv2.imread(osp.join('/cloud_disk/users/huh/pcb/script/slic_result/1_without_small',osp.basename(img_path)))#slic_image = cv2.resize(slic_image,(512,512))#remove_small_objects(mask_path,end_path,img_path,slic_image)except ValueError:print(end_path)

總結

以上是生活随笔為你收集整理的opencv实现超像素分割（slic实现）的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。