Convert PIL Image to byte array?

1.
import io

img = Image.open(fh, mode='r')
roiImg = img.crop(box)

imgByteArr = io.BytesIO()
roiImg.save(imgByteArr, format='PNG')
imgByteArr = imgByteArr.getvalue()

2.
from PIL import Image
import io

# I don't know what Python version you're using, so I'll try using Python 3 first
try:
    import urllib.request as urllib
except ImportError:
    # You are using Python 2 it turns out
    import urllib

def my_func(filename, ext):
    # Get the image from the URL
    im = Image.open(urllib.urlopen(filename))

    fp = io.BytesIO()
    format = Image.registered_extensions()['.'+ext]
    im.save(fp, format)
    return fp.getvalue()

jpg_bin = my_func("http://p1.pstatp.com/list/300x196/pgc-image/152923179745640a81b1fdc.webp", "jpg")

3.
import io
from PIL import Image  # 注意我的Image版本是pip3 install Pillow==4.3.0
import requests

res = requests.get('http://images.xxx.com/-7c0dc4dbdca3.webp', stream=True)  # 获取字节流最好加stream这个参数,原因见requests官方文档

byte_stream = io.BytesIO(res.content)  # 把请求到的数据转换为Bytes字节流(这样解释不知道对不对，可以参照[廖雪峰](https://www.liaoxuefeng.com/wiki/0014316089557264a6b348958f449949df42a6d3a2e542c000/001431918785710e86a1a120ce04925bae155012c7fc71e000)的教程看一下)

roiImg = Image.open(byte_stream)   # Image打开Byte字节流数据

imgByteArr = io.BytesIO()     # 创建一个空的Bytes对象

roiImg.save(imgByteArr, format='PNG') # PNG就是图片格式，我试过换成JPG/jpg都不行

imgByteArr = imgByteArr.getvalue()   # 这个就是保存的图片字节流

# 下面这一步只是本地测试， 可以直接把imgByteArr，当成参数上传到七牛云
with open("./abc.png", "wb") as f:
    f.write(imgByteArr)