fpl2

f = torch.fft.fft2(image)
# 计算频率
freqs = torch.fft.fftfreq(image.shape[-1])
print(freqs)
# 设定阈值，用于分离高频和低频信息
threshold = 0.1
# 创建掩码，用于分离高频和低频信息
mask = (freqs.abs() < threshold).float()
# 应用掩码，分离高频和低频信息
low_freq = torch.fft.ifft2(f * mask)
high_freq = image - low_freq