小波变换图像风格迁移代码

以下是使用小波变换进行图像风格迁移的Python代码，使用的是基于PyWavelets库进行小波变换：

import cv2
import numpy as np
import pywt

# Load images
content_img = cv2.imread('content.jpg')
style_img = cv2.imread('style.jpg')

# Resize images to the same size
content_img = cv2.resize(content_img, (256, 256))
style_img = cv2.resize(style_img, (256, 256))

# Convert images to YCbCr color space
content_img_y = cv2.cvtColor(content_img, cv2.COLOR_BGR2YCR_CB)[:,:,0]
style_img_y = cv2.cvtColor(style_img, cv2.COLOR_BGR2YCR_CB)[:,:,0]

# Apply 2D discrete wavelet transform to content image
coeffs = pywt.dwt2(content_img_y, 'haar')
LL, (LH, HL, HH) = coeffs

# Apply 2D discrete wavelet transform to style image
coeffs = pywt.dwt2(style_img_y, 'haar')
LLs, (LHs, HLs, HHs) = coeffs

# Combine coefficients of content and style images
coeffs_combined = (LL, 0.1*LHs + 0.9*LH, 0.1*HLs + 0.9*HL, 0.1*HHs + 0.9*HH)

# Apply inverse 2D discrete wavelet transform to combined coefficients
img_out_y = pywt.idwt2(coeffs_combined, 'haar')

# Merge Y channel with Cb and Cr channels of original content image
img_out = content_img.copy()
img_out[:,:,0] = img_out_y
img_out = cv2.cvtColor(img_out, cv2.COLOR_YCR_CB2BGR)

# Save output image
cv2.imwrite('output.jpg', img_out)

这段代码使用haar小波作为小波基，将content和style图片转换为YCbCr颜色空间，将content图片的LL（低频）系数和style图片的LH、HL、HH（高频）系数合并，最后再进行逆小波变换，得到迁移后的图像。可以根据需要自行修改代码中的参数，如小波基、系数权重等。