大家好，我是阿潘，今天跟大家分享一个赛博朋克特效的实现算法， 源码放在下面了哈
原理部分： 帅爆! 赛博朋克特效实现
为了写好这篇文章, 找了很多资料, 如果对你有帮助的话, 帮忙分享一波, 支持一下, 感谢!!!
在上次的文章分享处理之后，也有小伙伴留言其他效果更好的 案例给我，我也试着去弄，但是效果并不理想！从我自己的尝试，得到的一些结论就是最好原图要色彩比较丰富的，另外代码中的一些参数可以试着调整一下可能有奇效！
https://www.bilibili.com/video/BV1HJ411U7Qu
原图：
效果图：
原图：
效果图：
import cv2
import numpy as np
def modify_color_temperature(img):
# ---------------- 冷色調 ---------------- #
# 1.计算三个通道的平均值，并依据平均值调整色调
imgB = img[:, :, 0]
imgG = img[:, :, 1]
imgR = img[:, :, 2]
# 调整色调 # 白平衡 -> 三个值变化相同
# 冷色调(增加b分量) -> 除了b之外都增加
# 暖色调(增加r分量) -> 除了r之外都增加
bAve = cv2.mean(imgB)[0]
gAve = cv2.mean(imgG)[0] 10
rAve = cv2.mean(imgR)[0] 10
aveGray = (int)(bAve gAve rAve) / 3
# 2. 计算各通道增益系数，并使用此系数计算結果
bCoef = aveGray / bAve
gCoef = aveGray / gAve
rCoef = aveGray / rAve
imgB = np.floor((imgB * bCoef)) # 向下取整
imgG = np.floor((imgG * gCoef))
imgR = np.floor((imgR * rCoef))
# 3. 变换后处理
imgb = imgB
imgb[imgb > 255] = 255
imgg = imgG
imgg[imgg > 255] = 255
imgr = imgR
imgr[imgr > 255] = 255
cold_rgb = np.dstack((imgb, imgg, imgr)).astype(np.uint8)
return cold_rgb
def reverse_hue(image):
# 反转色相
image_hls = cv2.cvtColor(image, cv2.COLOR_BGR2HLS)
image_hls = np.asarray(image_hls, np.float32)
hue = image_hls[:, :, 0]
hue[hue < 90] = 180 - hue[hue < 90] - 10
image_hls[:, :, 0] = hue
image_hls = np.asarray(image_hls, np.uint8)
image = cv2.cvtColor(image_hls, cv2.COLOR_HLS2BGR)
return image
def cyberpunk(image):
image_lab = cv2.cvtColor(image, cv2.COLOR_BGR2Lab)
image_lab = np.asarray(image_lab, np.float32)
image_lab[:,:,0] = np.clip(image_lab[:,:,0] * 1.2,0,255)
# 提高像素亮度，让亮的地方更亮
light_gamma_high = np.power(image_lab[:, :, 0], 0.9)
light_gamma_high = np.asarray(light_gamma_high / np.max(light_gamma_high) * 255, np.uint)
# 降低像素亮度，让暗的地方更暗
light_gamma_low = np.power(image_lab[:, :, 0], 1.1)
light_gamma_low = np.asarray(light_gamma_low / np.max(light_gamma_low) * 255, np.uint8)
# 调色至偏紫
dark_b = image_lab[:, :, 2] * (light_gamma_low / 255) * 0.4
dark_a = image_lab[:, :, 2] * (1 - light_gamma_high / 255) * 0.1
image_lab[:, :, 2] = np.clip(image_lab[:, :, 2] - dark_b, 0, 255)
image_lab[:, :, 1] = np.clip(image_lab[:, :, 1] - dark_a, 0, 255)
image_lab = np.asarray(image_lab, np.uint8)
return cv2.cvtColor(image_lab, cv2.COLOR_Lab2BGR)
if __name__ == "__main__":
# 设置窗口可缩放
cv2.namedWindow('origin', cv2.WINDOW_NORMAL | cv2.WINDOW_KEEPRATIO)
cv2.namedWindow('cold_style', cv2.WINDOW_NORMAL | cv2.WINDOW_KEEPRATIO)
cv2.namedWindow('reverser_hue', cv2.WINDOW_NORMAL | cv2.WINDOW_KEEPRATIO)
cv2.namedWindow('cyberpunk', cv2.WINDOW_NORMAL | cv2.WINDOW_KEEPRATIO)
image = cv2.imread("./city.png")
cv2.imshow("origin", image)
image = modify_color_temperature(image)
cv2.imshow("cold_style", image)
image = reverse_hue(image)
cv2.imshow("reverser_hue", image)
# cv2.waitKey()
image = cyberpunk(image)
cv2.imshow("cyberpunk", image)
cv2.imwrite("result.jpg", image)
cv2.waitKey()
参考资料：
https://www.jianshu.com/p/7058debaaf2f
https://en.wikipedia.org/wiki/Cyberpunk
https://www.zhihu.com/question/269194796/answer/652003989
https://blog.csdn.net/sinat_26917383/article/details/70860910
https://www.w3schools.com/colors/colors_hsl.asp
https://blog.csdn.net/lysslq2015/article/details/124987039
https://docs.microsoft.com/en-us/windows/win32/wcs/hls-color-spaces
https://blog.csdn.net/weixin_48623445/article/details/119395509
https://rodrigoberriel.com/2014/11/opencv-color-spaces-splitting-channels/
https://learnopencv.com/color-spaces-in-opencv-cpp-python/
https://github.com/howarder3/ironman2020_OpenCV_photoshop/blob/master/Day06_調整色調色溫白平衡_modify_color_temperature.ipynb
https://en.wikipedia.org/wiki/HSL_and_HSV#HSL_to_RGB
https://sensing.konicaminolta.asia/what-is-cie-1976-lab-color-space/
https://zhuanlan.zhihu.com/p/158524543
