python圖像填充與裁剪/resize的實現代碼

from math import ceil
from torchvision.transforms.functional import to_tensor, to_pil_image
from PIL import Image
import torch
# 填充到最接近base整數倍的長和寬圖像大小
def get_padding_pic_mask(origin_png, result_png, result_mask_png, rgb=[[130, 206, 255]], base=4):
    # C, H, W
    src = Image.open(origin_png)
    src = to_tensor(src)
    # print(src.shape)  # torch.Size([3, 800, 600])
    # channel: (R, G, B) / 255
    origin_h, origin_w = src.shape[1], src.shape[2]
    print('原圖像大小, height: {}, width: {}'.format(origin_h, origin_w))

    h = ceil(origin_h / base) * base
    w = ceil(origin_w / base) * base

    img = torch.ones(3, h, w)
    # 如果想要填充是黑色則註釋掉上一句，換下面這一句
    # img = torch.zeros(3, h, w)

    img[:, :origin_h, :origin_w] = src 
    # 保存填充後的圖片
    to_pil_image(img).save(result_png)

    # 處理一下mask
    mask = torch.tensor(rgb) / 255

    mask = mask.view(3, 1, 1).repeat(1, h, w)
    # 保存填充後的mask
    to_pil_image(mask).save(result_mask_png)
    

# 圖像輸出後我們需要clip一下
def clip_unpadding(input_png, output_png, origin_h, origin_w):
    # C, H, W
    img = Image.open(input_png)
    img = to_tensor(img)
    img = img[:, :origin_h, :origin_w]
    # 保存裁剪後的圖片
    to_pil_image(img).save(output_png)

if __name__ == '__main__':
    # origin_png = 'pic/pic.jpg'
    # result_png = 'pic/pic_padding.jpg'
    # result_mask_png = 'pic/mask_padding.jpg'
    # get_padding_pic_mask(origin_png, result_png, result_mask_png)

    input_png = 'pic/pic_padding.jpg'
    output_png = 'pic/pic_clip.jpg'
    # 原圖像大小, height: 567, width: 390
    clip_unpadding(input_png, output_png, 567, 390)

from PIL import Image
def resize_img(origin_png, resize_png, height, width):
    img = Image.open(origin_png)
    img = img.resize((width, height))
    img.save(resize_png)
if __name__ == '__main__':
    origin_png = 'pic/white.jpg'
    resize_png = 'pic/white_resize.png'
    resize_img(origin_png, resize_png, 800, 600)