使用 python, OpenCV 建立

目錄

• 1. import cv2
• 2. 讀取影像、調整大小
• 3. 顯示影像
• 4. 保存影像
• 5. 複製影像
• 6. 影像RGB分開顯示
• 7. 上下左右翻轉影像
• 8. 兩個影像重疊 + Trackbar
• 9. Smoothing 影像
• 10. 邊界偵測
• 11. 影像 Transform - rotation, scaling, Trans_x, Trans_y

1. import cv2

2. 讀取影像、調整大小

• 讀取影像

image = cv2.imread('影像名稱.jpg/jpeg/png/tiff...')

• 讀取大小

image_size = image.shape
image_size = cv2.GetSize(image)

• 調整大小

reimage = cv2.resize(image, (64,64), interpolation=cv2.INTER_CUBIC)
# cv2.resize(原影像, 調整後的(W, H), 模式)

interpolation模式：

1. cv2.INTER_NEAREST: 採用鄰近插植
2. cv2.INTER_LINEAR: 雙線性插植(如果未設定，通常系統會默認為此模式)
3. cv2.INTER_AREA: 使用像素區域關係進行重採樣
4. cv2.INTER_CUBIC: 4*4像素鄰域的雙/三次插值
5. cv2.INTER_LANCZOS4: 8*8像素鄰域的Lanczos 插值

3. 顯示影像

# 創建一個新的窗口用於顯示圖像
cv2.nameWindow('名稱')
# 在窗口中顯示圖像
cv2.imshow('名稱', image)
cv2.waitKey(0)

4. 保存影像

v2.imwrite('路徑', image, 格式)

5. 複製影像

img = image.copy()

6. 影像RGB分開顯示

(B, G, R) = cv2.split(image)
zeros = np.zeros(image.shape[:2], dtype="uint8")

cv2.namedWindow('B')
cv2.imshow('B', cv2.merge([B, zeros, zeros]))
cv2.namedWindow('G')
cv2.imshow('G', cv2.merge([zeros, G, zeros]))
cv2.namedWindow('R')
cv2.imshow('R', cv2.merge([zeros, zeros, R]))

cv2.waitKey(0)

7. 上下左右翻轉影像

# 橫向翻轉
horizontal = cv2.flip(image, 1)
# 縱向翻轉
vertical = cv2.flip(image, 0)
# 橫向縱向翻轉
hv_both = cv2.flip(image, -1)

8. 兩個影像重疊 + Trackbar

org_img = cv2.imread('./Dataset_opencvdl/Q1_Image/Uncle_Roger.jpg')
flip_img = cv2.flip(org_img, 1)

def on_trackbar(val):
    alpha = val / 255
    beta = (1.0 - alpha)
    dst = cv2.addWeighted(org_img, alpha, flip_img, beta, 0.0)
    cv2.imshow('BLENDING', dst)

cv2.namedWindow('BLENDING')
cv2.createTrackbar('BLEND', 'BLENDING', 0, 255, on_trackbar)
cv2.waitKey(0)
cv2.destroyAllWindows()

OpenCV官網Two Images。
OpenCV官網Trackbar。

9. Smoothing 影像

• Gaussian

# (3, 3)為filter
guassian = cv2.GaussianBlur(image, (3, 3), 0)

• Median

# 7為K值
median = cv2.medianBlur(image, 7)

• Bilateral

bilateral = cv2.bilateralFilter(image, 9, 90, 90)

可參考此網址。
OpenCV官網Smoothing Image。

10. 邊界偵測

• 先做 Gaussian轉換

image = cv2.imread('./Dataset_opencvdl/Q3_Image/Chihiro.jpg')
gray_img = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)

# 3*3 Gaussian Filter
x, y = np.mgrid[-1:2, -1:2]
gaussian_kernel = np.exp(-(x ** 2 + y ** 2))
# Normalization
gaussian_kernel = gaussian_kernel / gaussian_kernel.sum()

grad = signal.convolve2d(gray_img, gaussian_kernel, boundary='symm', mode='same')  # convolution

plt.imshow(grad, cmap=plt.get_cmap('gray'))
plt.axis('off')
plt.show()

參考網址。

• Sobel X

image = cv2.imread('./Dataset_opencvdl/Q3_Image/Chihiro.jpg')
gray_img = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
gray_img = np.array(gray_img)

img_X = np.zeros(gray_img.shape)
# Filter
filters.sobel(gray_img, 1, img_X)
# Normalization
img_X *= (img_X - np.min(img_X)) / (np.max(img_X) - np.min(img_X))

plt.imshow(img_X, cmap='gray')
plt.axis('off')
plt.show()

• Sobwl Y

image = cv2.imread('./Dataset_opencvdl/Q3_Image/Chihiro.jpg')
gray_img = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
gray_img = np.array(gray_img)

img_Y = np.zeros(gray_img.shape)
# Filter
filters.sobel(gray_img, 0, img_Y)
# Normalization
img_Y *= (img_Y - np.min(img_Y)) / (np.max(img_Y) - np.min(img_Y))

plt.imshow(img_Y, cmap='gray')
plt.axis('off')
plt.show()

• Magnitude

image = cv2.imread('./Dataset_opencvdl/Q3_Image/Chihiro.jpg')
gray_img = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
gray_img = np.array(gray_img)

img_X = np.zeros(gray_img.shape)
filters.sobel(gray_img, 1, img_X)
img_Y = np.zeros(gray_img.shape)
filters.sobel(gray_img, 0, img_Y)

# Filter
mag = np.hypot(img_X, img_Y)
# Normalization
mag *= 255.0 / np.max(mag)

plt.imshow(mag, cmap='gray')
plt.axis('off')
plt.show()

11. 影像 Transform - rotation, scaling, Trans_x, Trans_y

image = cv2.imread('./Dataset_opencvdl/Q4_Image/Parrot.png')  # image.shape=(600, 900, 3)

center = ((image.shape[1] // 2 + float(Tx)), (image.shape[0] // 2 - float(Ty)))
rot_mat = cv2.getRotationMatrix2D(center, float(rotation), float(scaling))
rot_dst = cv2.warpAffine(image, rot_mat, (image.shape[1], image.shape[0]))

cv2.namedWindow('Image RST')
cv2.imshow('Image RST', rot_dst)

參考網址。