Profile Image of the Author
逍遥之韶
欲穷千里目，更上一层楼。
1395 字
7 分钟
OpenCV案例：滤镜
2023-09-12
计算机视觉
OpenCV
/
数字图像处理

1. 介绍#

​ 滤镜是图像处理过程中常见的一种应用，它能够改变图像原有的颜色、轮廓等信息，使图像看起来更具艺术效果。

2. 颜色效果#

2.1 颜色通道提取#

import cv2


img = cv2.imread("family.jpg")


B = img[ : , : , 0]
G = img[ : , : , 1]
R = img[ : , : , 2]


cv2.imwrite("family-B.jpg", B)
cv2.imwrite("family-G.jpg", G)
cv2.imwrite("family-R.jpg", R)
原图蓝色通道绿色通道红色通道

2.2 红绿互换#

import cv2


img = cv2.imread("family.jpg")


B, G, R = cv2.split(img)


img = cv2.merge([B, R, G])


cv2.imwrite("family-BRG.jpg", img)
原图效果图

2.3 灰度化#

import cv2


BGR = cv2.imread("family.jpg")


GRAY = cv2.cvtColor(BGR, cv2.COLOR_BGR2GRAY)


cv2.imwrite("family-gray.jpg", GRAY)
原图效果图

2.4 二值化#

import cv2


BGR = cv2.imread("family.jpg")


GRAY = cv2.cvtColor(BGR, cv2.COLOR_BGR2GRAY)


val, BINARY = cv2.threshold(src=GRAY, thresh=127, maxval=255, type=cv2.THRESH_BINARY)


cv2.imwrite("family-binary.jpg", BINARY)
原图灰度图二值图

3. 光线效果#

3.1 反相#

import cv2


src = cv2.imread("family.jpg")


dst = 255 - src


cv2.imwrite("family-invert.jpg", dst)
原图效果图

3.2 黑白底片#

import cv2


src = cv2.imread("family.jpg")


dst = cv2.cvtColor(255 - src, cv2.COLOR_BGR2GRAY)


cv2.imwrite("family-invert-gray.jpg", dst)
原图效果图

3.3 强光#

强光计算公式

import cv2
import numpy as np


src = cv2.imread("family.jpg")
H, W, C = src.shape
dst = np.zeros(shape=(H, W, C), dtype="uint8")


for h in range(0, H):
    for w in range(0, W):
        for c in range(0, C):
            pixel = float(src[h,w,c])
            if pixel > 127.5:
                pixel = pixel + (255.0 - pixel) * (pixel - 127.5) / 127.5
                pixel = np.clip(a=pixel, a_min=0, a_max=255).astype("uint8")
                dst[h,w,c] = pixel
            else:
                pixel = pixel + pixel * (pixel - 127.5) / 127.5
                pixel = np.clip(a=pixel, a_min=0, a_max=255).astype("uint8")
                dst[h,w,c] = pixel


cv2.imwrite("family-glare.jpg", dst)
原图效果图

4. 艺术效果#

4.1 模糊#

​ 使用自定义滤波器，实现模糊效果。

import cv2
import numpy as np


src = cv2.imread("family.jpg")


kernel = [[1, 2, 1], [2, 4, 2], [1, 2, 1]]
kernel = np.array(kernel, np.float32) / np.sum(kernel)
dst = cv2.filter2D(src, -1, kernel)


cv2.imwrite("family-blur.jpg", dst)
原图效果图

4.2 柔和#

​ 使用3*3的均值滤波器，实现柔和效果。

import cv2


src = cv2.imread("family.jpg")


dst = cv2.blur(src, (3, 3))


cv2.imwrite("family-soft.jpg", dst)
原图效果图

4.3 锐化#

​ 首先计算当前像素点与周围像素点平均值之差，再将该差值与当前点进行算术运算，求得当前像素点的锐化值。

  • 差值 = 当前像素点 - 周围像素点的平均值
  • 锐化像素点 = 当前像素点 + 差值 * 锐化系数
import cv2
import numpy as np


src = cv2.imread("family.jpg")
H, W, C = src.shape
dst = np.zeros(shape=(H, W, C), dtype="uint8")


coff = 2
src = src.astype(np.float32)
for h in range(1, H-1):
    for w in range(1, W-1):
        for c in range(0, C):
            LU = src[h-1,w-1,c]
            MU = src[h-1, w ,c]
            RU = src[h-1,w+1,c]
            L  = src[ h ,w-1,c]
            R  = src[ h ,w+1,c]
            LB = src[h+1,w-1,c]
            MB = src[h+1, w ,c]
            RB = src[h+1,w+1,c]
            MEAN = np.mean([LU, MU, RU, L, R, LB, MB, RB])
            SHARP = src[h,w,c] + (src[h,w,c] - MEAN) * coff
            dst[h,w,c] = np.clip(SHARP, a_min=0, a_max=255).astype(np.uint8)


cv2.imwrite("family-sharp.jpg", dst)
原图效果图

4.4 扩散#

​ 扩散指用当前像素点或其周边的像素点，替换当前像素点。选择逻辑是随机选择。

import cv2
import numpy as np


src = cv2.imread("family.jpg")
H, W, C = src.shape
dst = np.zeros(shape=(H, W, C), dtype="uint8")


for h in range(1, H-1):
    for w in range(1, W-1):
        for c in range(0, C):
            LU = src[h-1,w-1,c]
            MU = src[h-1, w ,c]
            RU = src[h-1,w+1,c]
            L  = src[ h ,w-1,c]
            M  = src[ h , w ,c]
            R  = src[ h ,w+1,c]
            LB = src[h+1,w-1,c]
            MB = src[h+1, w ,c]
            RB = src[h+1,w+1,c]
            LIST = np.array([LU, MU, RU, L, M, R, LB, MB, RB])
            index = np.random.randint(0, len(LIST))
            dst[h,w,c] = condition[index]


cv2.imwrite("family-diffusion.jpg", dst)
原图效果图

4.5 老照片#

​ 传统的纸质相片在经过岁月的洗礼后会泛黄，在图像上表现为黄褐色效果。

import cv2
import numpy as np


src = cv2.imread("family.jpg")
dst = np.zeros(shape=(src.shape), dtype="uint8")


src = src.astype(np.uint32)
B = (34 * src[:,:,0] + 137 * src[:,:,1] +  70 * src[:,:,2]) // 256
G = (43 * src[:,:,0] + 176 * src[:,:,1] +  89 * src[:,:,2]) // 256
R = (48 * src[:,:,0] + 197 * src[:,:,1] + 101 * src[:,:,2]) // 256


dst[:,:,0] = np.clip(B, a_min=0, a_max=255).astype(np.uint8)
dst[:,:,1] = np.clip(G, a_min=0, a_max=255).astype(np.uint8)
dst[:,:,2] = np.clip(R, a_min=0, a_max=255).astype(np.uint8)


cv2.imwrite("family-old-photo.jpg", dst)
原图效果图

4.6 木雕画#

​ 木雕画指将图像雕刻在木雕上。将图像处理为木雕画效果，可以将图像的轮廓以白色显示，背景用黑色显示。

import cv2
import numpy as np


src = cv2.imread("family.jpg", 0)
H, W = src.shape
dst = np.zeros(shape=(H, W), dtype="uint8")


delta = 64
src = src.astype(np.float32)
for h in range(1, H-1):
    for w in range(1, W-1):
        LU = np.abs(src[h,w] - src[h-1,w-1])
        MU = np.abs(src[h,w] - src[h-1, w ])
        RU = np.abs(src[h,w] - src[h-1,w+1])
        L  = np.abs(src[h,w] - src[ h ,w-1])
        M  = np.abs(src[h,w] - src[ h , w ])
        R  = np.abs(src[h,w] - src[ h ,w+1])
        LB = np.abs(src[h,w] - src[h+1,w-1])
        MB = np.abs(src[h,w] - src[h+1, w ])
        RB = np.abs(src[h,w] - src[h+1,w+1])
        condition = np.array([LU, MU, RU, L, M, R, LB, MB, RB])
        if (any(condition>delta)):
            dst[h,w] = 255
        else:
            dst[h,w] = 0


cv2.imwrite("family-wooden-sculpture.jpg", dst)
原图效果图

4.7 铅笔画#

​ 铅笔画指用铅笔在白纸上作画。将图像处理为铅笔画效果，可以将图像的轮廓以黑色显示，背景用白色显示。

import cv2
import numpy as np


src = cv2.imread("family.jpg", 0)
H, W = src.shape
dst = np.zeros(shape=(H, W), dtype="uint8")


delta = 40
src = src.astype(np.float32)
for h in range(1, H-1):
    for w in range(1, W-1):
        LU = np.abs(src[h,w] - src[h-1,w-1])
        MU = np.abs(src[h,w] - src[h-1, w ])
        RU = np.abs(src[h,w] - src[h-1,w+1])
        L  = np.abs(src[h,w] - src[ h ,w-1])
        M  = np.abs(src[h,w] - src[ h , w ])
        R  = np.abs(src[h,w] - src[ h ,w+1])
        LB = np.abs(src[h,w] - src[h+1,w-1])
        MB = np.abs(src[h,w] - src[h+1, w ])
        RB = np.abs(src[h,w] - src[h+1,w+1])
        condition = np.array([LU, MU, RU, L, M, R, LB, MB, RB])
        if (any(condition>delta)):
            dst[h,w] = 0
        else:
            dst[h,w] = 255


cv2.imwrite("family-pencil.jpg", dst)
原图效果图

4.8 凸雕、阳刻#

​ 实现凸雕效果时，使用右下角像素点的值减去左上角像素点的值。

import cv2
import numpy as np


depth = 100


src = cv2.imread("family.jpg", 0)
H, W = src.shape
dst = np.zeros(shape=(H, W), dtype="uint8")


src = src.astype(np.float32)
for h in range(1, H-1):
    for w in range(1, W-1):
        LU = src[h-1,w-1]
        RB = src[h+1,w+1]
        pixel = RB - LU + depth
        dst[h,w] = np.clip(pixel, a_min=0, a_max=255).astype("uint8")


cv2.imwrite("family-embossment.jpg", dst)
原图效果图

4.9 凹雕、阴刻#

​ 实现凹雕效果时，使用左上角像素点的值减去右下角像素点的值。

import cv2
import numpy as np


depth = 100


src = cv2.imread("family.jpg", 0)
H, W = src.shape
dst = np.zeros(shape=(H, W), dtype="uint8")


src = src.astype(np.float32)
for h in range(1, H-1):
    for w in range(1, W-1):
        LU = src[h-1,w-1]
        RB = src[h+1,w+1]
        pixel = LU - RB + depth
        dst[h,w] = np.clip(pixel, a_min=0, a_max=255).astype("uint8")


cv2.imwrite("family-intaglio.jpg", dst)
原图效果图
OpenCV案例：手写数字识别
OpenCV案例：数字水印
© 2025 逍遥之韶. All Rights Reserved. / RSS / Sitemap
Powered by Astro & Mizuki  Version 4.9.5
封面
示例歌曲
示例艺术家
封面
示例歌曲
示例艺术家
0:00 / 0:00