3-1 (1) 感觉就是图像模糊了，并没有去噪
from cv2 import cv2 import numpy as np import random def spNoise(img,prob): # 添加椒盐噪声,prob:噪声比例 output = np.zeros(img.shape,np.uint8) thres = 1 - prob for i in range(img.shape[0]): for j in range(img.shape[1]): rdn = random.random() if rdn < prob: output[i][j] = 0 elif rdn > thres: output[i][j] = 255 else: output[i][j] = img[i][j] return output def gaussNoise(img, mean=0, var=0.0001): # 添加,高斯噪声mean : 均值,var : 方差 img = np.array(img/255, dtype=float) noise = np.random.normal(mean, var ** 0.5, img.shape) out = img + noise if out.min() < 0: low_clip = -1. else: low_clip = 0. out = np.clip(out, low_clip, 1.0) out = np.uint8(out*255) return out def meanFilter(img, c): # 均值滤波，2c+1*2c+1矩阵取平均值 img_shape = np.shape(img) out = np.zeros(img_shape) for i in range(img_shape[0]): for j in range(img_shape[1]): if i >= c and i < img_shape[0] - c and j >= c and j < img_shape[1] - c: out[i][j] = min(sum(img[i - c:i + c + 1, j - c:j + c + 1].flatten()) // ((2 * c + 1) * (2 * c + 1)), img[i][j]) else: out[i][j] = img[i][j] return out def medianFilter(img, c): # 中值滤波，2c+1*2c+1矩阵再取中值 img_shape = np.shape(img) out = np.zeros(img_shape) for i in range(img_shape[0]): for j in range(img_shape[1]): if i >= c and i < img_shape[0] - c and j >= c and j < img_shape[1] - c: out[i][j] = medianValueOdd(img[i - c:i + c + 1, j - c:j + c + 1].flatten()) else: out[i][j] = img[i][j] return out def medianValueOdd(arr): # 希尔排序 length = len(arr) gap = length//2 while gap > 0: for i in range(gap, length): tem = arr[i] j = i while j >= gap and arr[j-gap] >tem: arr[i], arr[i - gap] = arr[i - gap], arr[i] j -= gap gap = gap // 2 return arr[length//2] img = cv2.imread("C:\\test\\1.jpg") gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) img_noise = spNoise(gray, 0.0005) img_noise = gaussNoise(img_noise) img_mean_filter1 = meanFilter(img_noise, 1) img_mean_filter2 = meanFilter(img_noise, 2) img_mean_filter3 = meanFilter(img_noise, 3) img_median_filter1 = medianFilter(img_noise, 1) img_median_filter2 = medianFilter(img_noise, 2) img_median_filter3 = medianFilter(img_noise, 3) cv2.imwrite("C:\\test\\img_mean_filter1.jpg", img_mean_filter1) cv2.imwrite("C:\\test\\img_mean_filter2.jpg", img_mean_filter2) cv2.imwrite("C:\\test\\img_mean_filter3.jpg", img_mean_filter3) cv2.imwrite("C:\\test\\img_median_filter1.jpg", img_median_filter1) cv2.imwrite("C:\\test\\img_median_filter2.jpg", img_median_filter2) cv2.imwrite("C:\\test\\img_median_filter3.jpg", img_median_filter3) img_mean_filter1 = cv2.imread("C:\\test\\img_mean_filter1.jpg") img_mean_filter2 = cv2.imread("C:\\test\\img_mean_filter2.jpg") img_mean_filter3 = cv2.imread("C:\\test\\img_mean_filter3.jpg") img_median_filter1 = cv2.imread("C:\\test\\img_median_filter1.jpg") img_median_filter2 = cv2.imread("C:\\test\\img_median_filter2.jpg") img_median_filter3 = cv2.imread("C:\\test\\img_median_filter3.jpg") cv2.imshow('gray', gray) cv2.imshow('img_noise', img_noise) cv2.imshow('img_mean_filter1', img_mean_filter1) cv2.imshow('img_mean_filter2', img_mean_filter2) cv2.imshow('img_mean_filter3', img_mean_filter3) cv2.imshow('img_median_filter1', img_median_filter1) cv2.imshow('img_median_filter2', img_median_filter2) cv2.imshow('img_median_filter3', img_median_filter3) cv2.waitKey() cv2.destroyAllWindows4-1 (2)使用K邻近滤波和对称邻近滤波，发现K邻近滤波时，若仅取K=3，会产生很多早点，取K=5，K=7时，能有效去噪
from cv2 import cv2 import numpy as np import random def spNoise(img,prob): # 添加椒盐噪声,prob:噪声比例 output = np.zeros(img.shape,np.uint8) thres = 1 - prob for i in range(img.shape[0]): for j in range(img.shape[1]): rdn = random.random() if rdn < prob: output[i][j] = 0 elif rdn > thres: output[i][j] = 255 else: output[i][j] = img[i][j] return output def gaussNoise(img, mean=0, var=0.0001): # 添加,高斯噪声mean : 均值,var : 方差 img = np.array(img/255, dtype=float) noise = np.random.normal(mean, var ** 0.5, img.shape) out = img + noise if out.min() < 0: low_clip = -1. else: low_clip = 0. out = np.clip(out, low_clip, 1.0) out = np.uint8(out*255) return out def shellSort(arr): # 希尔排序 length = len(arr) gap = length//2 while gap > 0: for i in range(gap, length): tem = arr[i] j = i while j >= gap and arr[j-gap] >tem: arr[i], arr[i - gap] = arr[i - gap], arr[i] j -= gap gap = gap // 2 return arr def kNeighborFilter(img, N): square = N * N K = 2 * N - 1 q = N // 2 img_shape = np.shape(img) out = np.zeros(img_shape) for i in range(img_shape[0]): for j in range(img_shape[1]): if i >= q and i < img_shape[0] - q and j >= q and j < img_shape[1] - q: img_flatten = img[i - q:i + q + 1, j - q:j + q + 1].flatten() arr = np.append(img_flatten[0:square // 2], (img_flatten[square // 2 + 1:])) arr = shellSort(arr) p = 0 while p < len(arr)-1: if arr[p] > img[i][j]: break if arr[p] <= img[i][j] and arr[p + 1] >= img[i][j]: break p += 1 if p < K // 2: out[i][j] = sum(arr[0:K]) // K elif p > square - K // 2: out[i][j] = sum(arr[square - K:]) // K else: out[i][j] = sum(arr[p - K // 2:p + K // 2 + 1]) // K else: out[i][j] = img[i][j] return out def symmetricNeighborFilter(img, N): p = N // 2 img_shape = np.shape(img) out = np.zeros(img_shape) for i in range(img_shape[0]): for j in range(img_shape[1]): if i >= p and i < img_shape[0] - p and j >= p and j < img_shape[1] - p: tem = [] for k in range(i - p, i + p + 1): for l in range(j - p, j): tem.append([img[k][l], img[2 * i - k][2 * j - l]]) for k in range(i - p, i): tem.append([img[k, j], img[2 * i - k][j]]) total = 0 print(tem) for m in range(len(tem)): if abs(tem[m][0] - img[i][j]) < abs(tem[m][1] - img[i][j]): total += tem[m][0] else: total += tem[m][1] print(total) out[i][j] = total // len(tem) print(out[i][j]) else: out[i][j] = img[i][j] return out img = cv2.imread("C:\\test\\1.jpg") gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) img_noise = spNoise(gray, 0.0005) img_noise = gaussNoise(img_noise) k_neighbor_filter1 = kNeighborFilter(img_noise, 3) k_neighbor_filter2 = kNeighborFilter(img_noise, 5) k_neighbor_filter3 = kNeighborFilter(img_noise, 7) symmetric_neighbor_filter1 = symmetricNeighborFilter(img_noise, 3) symmetric_neighbor_filter2 = symmetricNeighborFilter(img_noise, 5) symmetric_neighbor_filter3 = symmetricNeighborFilter(img_noise, 7) cv2.imwrite("C:\\test\\kNeighborFilter3.jpg", k_neighbor_filter1) cv2.imwrite("C:\\test\\kNeighborFilter5.jpg", k_neighbor_filter2) cv2.imwrite("C:\\test\\kNeighborFilter7.jpg", k_neighbor_filter3) cv2.imwrite("C:\\test\\symmetricNeighborFilter3.jpg", symmetric_neighbor_filter1) cv2.imwrite("C:\\test\\symmetricNeighborFilter5.jpg", symmetric_neighbor_filter2) cv2.imwrite("C:\\test\\symmetricNeighborFilter7.jpg", symmetric_neighbor_filter3) k_neighbor_filter1 = cv2.imread("C:\\test\\kNeighborFilter3.jpg") k_neighbor_filter2 = cv2.imread("C:\\test\\kNeighborFilter5.jpg") k_neighbor_filter3 = cv2.imread("C:\\test\\kNeighborFilter7.jpg") symmetric_neighbor_filter1 = cv2.imread("C:\\test\\symmetricNeighborFilter3.jpg") symmetric_neighbor_filter2 = cv2.imread("C:\\test\\symmetricNeighborFilter5.jpg") symmetric_neighbor_filter3 = cv2.imread("C:\\test\\symmetricNeighborFilter7.jpg") cv2.imshow('gray', gray) cv2.imshow('img_noise', img_noise) cv2.imshow('kNeighborFilter3', k_neighbor_filter1) cv2.imshow('kNeighborFilter5', k_neighbor_filter2) cv2.imshow('kNeighborFilter7', k_neighbor_filter3) cv2.imshow('kNeighborFilter3', symmetric_neighbor_filter1) cv2.imshow('kNeighborFilter5', symmetric_neighbor_filter2) cv2.imshow('kNeighborFilter7', symmetric_neighbor_filter3) cv2.waitKey() cv2.destroyAllWindows4-1(3)看起来开运算去掉了白点，闭运算去掉了黑点
from cv2 import cv2 import numpy as np import random def spNoise(img,prob): # 添加椒盐噪声,prob:噪声比例 output = np.zeros(img.shape,np.uint8) thres = 1 - prob for i in range(img.shape[0]): for j in range(img.shape[1]): rdn = random.random() if rdn < prob: output[i][j] = 0 elif rdn > thres: output[i][j] = 255 else: output[i][j] = img[i][j] return output def gaussNoise(img, mean=0, var=0.0001): # 添加,高斯噪声mean : 均值,var : 方差 img = np.array(img/255, dtype=float) noise = np.random.normal(mean, var ** 0.5, img.shape) out = img + noise if out.min() < 0: low_clip = -1. else: low_clip = 0. out = np.clip(out, low_clip, 1.0) out = np.uint8(out*255) return out def grayErosion(img, N): arr = np.ones((N, N)) p = N // 2 img_shape = np.shape(img) out = np.zeros(img_shape) for i in range(img_shape[0]): for j in range(img_shape[1]): if i >= p and i < img_shape[0] - p and j >= p and j < img_shape[1] - p: out[i][j] = max(0, min(img[i - p:i + p + 1, j - p:j + p + 1].flatten()-arr.flatten())) else: out[i][j] = img[i][j] return out def grayScale(img, N): arr = np.ones((N, N)) p = N // 2 img_shape = np.shape(img) out = np.zeros(img_shape) for i in range(img_shape[0]): for j in range(img_shape[1]): if i >= p and i < img_shape[0] - p and j >= p and j < img_shape[1] - p: out[i][j] = min(255, max(img[i - p:i + p + 1, j - p:j + p + 1].flatten()-arr.flatten())) else: out[i][j] = img[i][j] return out def openFilter(img, arr): return grayScale(grayErosion(img, arr), arr) def closeFilter(img, arr): return grayErosion(grayScale(img, arr), arr) img = cv2.imread("C:\\test\\1.jpg") gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) img_noise = spNoise(gray, 0.0005) img_noise = gaussNoise(img_noise) image_open_filter = openFilter(img_noise, 3) image_close_filter = closeFilter(img_noise, 3) cv2.imwrite("C:\\test\\image_open_filter.jpg", image_open_filter) cv2.imwrite("C:\\test\\image_close_filter.jpg", image_close_filter) image_open_filter = cv2.imread("C:\\test\\image_open_filter.jpg") image_close_filter = cv2.imread("C:\\test\\image_close_filter.jpg") cv2.imshow('gray', gray) cv2.imshow('img_noise', img_noise) cv2.imshow('image_open_filter', image_open_filter) cv2.imshow('image_close_filter', image_close_filter) cv2.waitKey() cv2.destroyAllWindows