深度学习与图像处理之：人像背景虚化

简单实现思路：python

1. 对图像内容进行分割，提取人像
2. 对图像背景进行模糊化处理
3. 将人像和背景从新合成

在这里，使用DeepLabV3模型对图像内容进行分割并提取人像，实现的代码以下：优化

import numpy as np
import tensorflow as tf
import cv2
from deeplabmodel import *

def create_pascal_label_colormap():
    colormap = np.zeros((256, 3), dtype=int)
    ind = np.arange(256, dtype=int)

    for shift in reversed(range(8)):
        for channel in range(3):
            colormap[:, channel] |= ((ind >> channel) & 1) << shift
            ind >>= 3
    return colormap

def label_to_color_image(label):
    if label.ndim != 2:
        raise ValueError('Expect 2-D input label')

    colormap = create_pascal_label_colormap()

    if np.max(label) >= len(colormap):
        raise ValueError('label value too large.')
    return colormap[label]

def load_model():
    model_path = '../resources/models/tensorflow/deeplabv3_mnv2_pascal_train_aug_2018_01_29.tar.gz'#'deeplab_model.tar.gz'
    MODEL = DeepLabModel(model_path)
    print('model loaded successfully!')
    return MODEL

model = load_model()

src = cv2.imread('../resources/images/person2.jpg')
resized_im, seg_map = model.run2(src)
seg_image = label_to_color_image(seg_map).astype(np.uint8)
print(seg_map.dtype)
# seg_map = cv2.GaussianBlur(np.uint8(seg_map),(11,11),0)
src_resized = cv2.resize(src,(resized_im.shape[1],resized_im.shape[0]))
seg_image = cv2.GaussianBlur(seg_image,(11,11),0)
bg_img = np.zeros_like(src_resized)

bg_img[seg_map == 0] = src_resized[seg_map == 0]

blured_bg = cv2.GaussianBlur(bg_img,(11,11),0)
result = np.zeros_like(bg_img)

result[seg_map > 0] = resized_im[seg_map > 0]
result[seg_map == 0] = blured_bg[seg_map == 0]

cv2.imshow("seg_image",seg_image)
cv2.imshow('bg_image',bg_img)
cv2.imshow('blured_bg',blured_bg)
cv2.imshow('result',result)


cv2.waitKey()
cv2.destroyAllWindows()

原图：ui

人像提取结果：code

背景图像：orm

背景模糊图像：blog

合成结果：input

效果不太理想，但整体上实现了背景虚化。后期将进行细节优化。it