图像数据处理–Pillow

1.Pillow 概览

PIL: Python Imaging Library，Python 图像处理库，提供了如下的功能:

  • Image Archives, 图像存档和批处理

    • 图像读取(read image)
    • 图像打印(print image)
    • 图像缩略图(create thumbnails)
    • 图像格式变换(convert file format)

  • Image Display, 图像显示

    • TK PhotoImage
    • TK BitmapImage
    • Windows DIB interface[PythonWin, Windows-based tookits]
    • show[debugging]

  • Image Processing, 图像处理

    • 图像点操作(point operations)
    • 图像卷积核滤波(filtering with a set of build in convolution kernel)
    • 图像颜色空间转换(image colour space conversions)
    • 图像缩放(image resizing)
    • 图像旋转(image rotation)
    • 图像仿射变换(image arbitray affine transforms)
    • 图像统计信息分析直方图(histogram method)

2.Pillow 安装

  • macOS 安装:

    $ python3 -m pip install --upgrade pip
$ python3 -m pip install --upgrade Pillow

  • Windows 安装:

    $ python3 -m pip install --upgrade pip
$ python3 -m pip install --upgrade Pillow

  • Linux 安装:

    $ python3 -m pip install --upgrade pip
$ python3 -m pip install --upgrade Pillow

3.Pillow 核心

3.1 Image class

API:

  • from PIL import Image
  • Image.open
  • Image.format
  • Image.size
  • Image.mode
  • Image.show

示例:

import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"
try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:

        # 图像格式
        print(im.format)

        # 图像尺寸
        print(im.size)

        # 图像模式
        print(im.mode)

        # 图像打印
        im.show()
except IOError as e:
    print(f"Can't open {image_name}")
../_images/lena.png

Note

  • .format: 图像格式
  • .size: 图像尺寸(width_pixels, height_pixels)
  • .mode: 图像中条带的数量和名称、像素类型、像素深度

3.2 图像读、写、转换

API:

  • Image.open
  • Image.save
  • Image.thumbnail(size = (width, height))

示例:

  • 将图像转换为 JPEG 格式:

    import os, sys
from PIL import Image

for infile in sys.argv[1:]:
    f, e = os.path.splitext(inflie)
    outfile = f + ".jpg"
    if infile != outfile:
        try:
            with Image.open(infile) as im:
                im.save(outfile)
        except OSError:
            print("cannot convert", infile)

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"
try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        im.save(os.path.join(image_path, image_name + ".jpg"))
except IOError as e:
    print(f"Can't open {image_name}")
    ../_images/lena.jpg

  • 创建 JPEG 缩略图(thumbnails)

    import os, sys
from PIL import Image

size = (128, 128)

for inflie in sys.argv[1:]:
    outfile = os.path.splitext(infile)[0] + ".thumbnail"
    if infile != outfile:
        try:
            with Image.open(infile) as im:
                im.thumbnail(size)
                im.save(outfile, "JPEG")
        except OSError:
            print(f"Can't create thumbnail for{infile}")

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"
size = (128, 128)

try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        im.thumbnail(size)
        im.save(os.path.join(image_path, image_name + ".JPEG"))
except IOError as e:
    print(f"Can't open {image_name}")
    ../_images/lena.JPEG

  • 识别图像文件

    import sys
from PIL import Image

for infile in sys.argv[1:]:
    try:
        with Image.open(infile) as im:
            print(infile, im.format, f"{im.size}x{im.mode}")
    except OSError:
        pass

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"

try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        print(image_name, im.format, f"{im.size}x{im.mode}")
except IOError as e:
    print(f"Can't open {image_name}")

3.3 图像剪切、粘贴、拼接

API:

  • Image.crop: 从图像中复制子矩形
  • Image.past
  • Image.merge
  • Image.split
  • Image.transpose(Image.ROTATE_180)

示例:

  • 从图像中复制子矩形

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"

try:
    box = (100, 100, 200, 200)
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        region = im.crop(box)
        region.show()
        region.save(os.path.join(image_path, image_name + "_region" + ".png"))
except IOError as e:
    print(f"Can't open {image_name}")
    ../_images/lena_region.png

    Note

    • PIL 中图像左上角坐标为 (0, 0)
    • box(left, upper, right, lowe)

  • 从图像中复制子矩形、将子矩形粘贴回去

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"

try:
    box = (100, 100, 200, 200)
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        region = im.crop(box)
        region = region.transpose(Image.ROTATE_180)
        im.paste(region, box)
        im.save(os.path.join(image_path, image_name + "_region_paste" + ".png"))
except IOError as e:
    print(f"Can't open {image_name}")
    ../_images/lena_region_paste.png

  • 图像滚动(image roll)

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"

def roll(image, delta):
    """Roll an image sideways"""
    xsize, ysize = image.size
    delta = delta % xsize
    if delta == 0:
        return image
    part1 = image.crop((0, 0, delta, ysize))
    part2 = image.crop((delta, 0, xsize, ysize))
    image.paste(part1, (xsize - delta, 0, xsize, ysize))
    image.paste(part2, (0, 0, xsize - delta, ysize))

    return image

try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        im = roll(im, 10)
        im.save(os.path.join(image_path, image_name + "_roll" + ".png"))
except OSError:
    print(f"cannot open {image_name}")
    ../_images/lena_roll.png

  • RGB波段拆分、合并

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"

try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        r, g, b = im.split()
        im = Image.merge("RGB", (b, g, r))
        im.save(os.path.join(image_path, image_name + "_merge_gbr" + ".png"))
except OSError:
    print(f"cannot open {image_name}")
    ../_images/lena.png ../_images/lena_merge_rbg.png ../_images/lena_merge_brg.png ../_images/lena_merge_bgr.png ../_images/lena_merge_grb.png ../_images/lena_merge_gbr.png

    Note

    • 对于单波段图像(single-band)，Image.split 返回图像本身
    • 为了对单个颜色波段进行处理，需要首先将图像转换为 RGB

3.4 图像几何变换

API:

  • Image.Image.resize
  • Image.Image.rotate
  • Image.transpose
  • Image.transform

示例:

  • 简单的几何变换-改变图像像素大小

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"

try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        out = im.resize((1000, 1000))
        out.save(os.path.join(image_path, image_name + "_resize" + ".png"))
except OSError:
    print(f"cannot open {image_name}")
    ../_images/lena_resize.png

  • 简单的几何变换-图像逆时针旋转一定的角度

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"

try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        out = im.rotate(45)
        out.save(os.path.join(image_path, image_name + "_rotate" + ".png"))
except OSError:
    print(f"cannot open {image_name}")
    ../_images/lena_rotate.png

  • 图像转置

    import os
from PIL import Image

image_path = "/Users/zfwang/project/machinelearning/deeplearning/deeplearning/src/pillow_src/images"
image_name = "lena"

try:
    with Image.open(os.path.join(image_path, image_name + ".png")) as im:
        out1 = im.transpose(Image.FLIP_LEFT_RIGHT)
        out2 = im.transpose(Image.FLIP_TOP_BOTTOM)
        out3 = im.transpose(Image.ROTATE_90)
        out4 = im.transpose(Image.ROTATE_180)
        out5 = im.transpose(Image.ROTATE_270)
        out1.save(os.path.join(image_path, image_name + "_rotate_1" + ".png"))
        out2.save(os.path.join(image_path, image_name + "_rotate_2" + ".png"))
        out3.save(os.path.join(image_path, image_name + "_rotate_3" + ".png"))
        out4.save(os.path.join(image_path, image_name + "_rotate_4" + ".png"))
        out5.save(os.path.join(image_path, image_name + "_rotate_5" + ".png"))
except OSError:
    print(f"cannot open {image_name}")
    ../_images/lena.png ../_images/lena_rotate_1.png ../_images/lena_rotate_2.png ../_images/lena_rotate_3.png ../_images/lena_rotate_4.png ../_images/lena_rotate_5.png

Note

  • trasnpose(ROTATE) 与 Image.Image.rotate 效果相同
  • transform() 能进行更多形式的图像转换

3.5 图像颜色转换

3.6 图像增强

3.7 图像序列

3.8 PostScript 打印

3.9 图像读取

3.10 控制图像编码

4.Pillow API

  • Module

    • Image
    • ImageChops
    • ImageCms
    • ImageColor
    • ImageDraw
    • ImageEnhance
    • ImageFile
    • ImageFilter
    • ImageFont
    • ImageGrab
    • ImageMath
    • ImageMorph
    • ImageOps
    • ImagePaletee
    • ImagePath
    • ImageQt
    • ImageSequence
    • ImageShow
    • ImageStat
    • ImageTK
    • ImageWin
    • ExifTags
    • TiffTags
    • JpegPressets
    • PSDraw
    • PixelAccess
    • PyAccess
    • features

  • PIL Package

    • PIL

  • Plugin reference

  • Internal Reference Docs