s1081440作業4

題目敘述: 影像還原練習

附件中的image4 似乎受到某種頻域雜訊干擾，撰寫一個程式嘗試復原此圖像(將圖中雜訊去除)。

開發環境

Windows 11

VSCode  Python 3.10.4 + OpenCV 4.7.0

程式碼說明

import cv2

import numpy as np

import matplotlib.pyplot as plt

# 灰階讀取圖片

img = cv2.imread('image4.png', cv2.IMREAD_GRAYSCALE)

# Fast Fourier Transform 取得頻域圖

dft = cv2.dft(np.float32(img), flags=cv2.DFT_COMPLEX_OUTPUT)

dft_shift = np.fft.fftshift(dft)

freq_image = 20*np.log(cv2.magnitude(dft_shift[:, :, 0], dft_shift[:, :, 1]))

# 利用threshhold找出高頻干擾的位置

# (794, 377) (805, 411) (788, 441) (810, 509) (790, 542) (805, 573)

noise_img = cv2.threshold(freq_image, 250, 255, cv2.THRESH_BINARY)

# 製作mask

noise_mask = np.ones(dft_shift.shape, np.uint8)

noise_mask = cv2.circle(noise_mask, (794, 377), 10, (0, 0, 0), -1)

noise_mask = cv2.circle(noise_mask, (805, 411), 10, (0, 0, 0), -1)

noise_mask = cv2.circle(noise_mask, (788, 441), 10, (0, 0, 0), -1)

noise_mask = cv2.circle(noise_mask, (810, 509), 10, (0, 0, 0), -1)

noise_mask = cv2.circle(noise_mask, (790, 542), 10, (0, 0, 0), -1)

noise_mask = cv2.circle(noise_mask, (805, 573), 10, (0, 0, 0), -1)

# Inverse Fast Fourier Transform

fshift = dft_shift*noise_mask

ishift = np.fft.ifftshift(fshift)

iimg = cv2.idft(ishift)

img_back = cv2.magnitude(iimg[:, :, 0], iimg[:, :, 1])

# show result

plt.imshow(img_back, 'gray')

plt.axis('off')

plt.show()

找出的干擾處:

根據下圖找到的6的光點干擾製作mask與原dft圖相乘(fshift = dft_shift*noise_mask)，遮掉干擾後印出結果

執行結果