不幸的是，我无法阅读你提供的文件，因为网站已经将其大量拆分。使用pastebin或github（或类似的工具），可以检索text/plain，例如通过curl，这样我就可以尝试1：1重现问题的内容。
但是，一般的做法是：

from PIL import Image

with Image.open("./test_image_3.txt") as im:
    im.show()

它直接来自Pillow的documentation，不关心文件名或扩展名。
或者，如果您使用文件句柄调用open()：

from PIL import Image

with open("./test_image_3.txt", "rb") as file:
    with Image.open(file) as im:
        im.show()

如果它被损坏了，那么从encode()和decode()调用判断，它将是这样的：

from PIL  import Image
from io import BytesIO

data = <some raw PNG bytes, the original image>

# here I store it in that weird format and write as bytes
with open("img.txt", "wb") as file:
    file.write(data.decode("latin-1").encode("unicode_escape"))

# here I read it back as bytes, reverse the chain of calls and invert
# the call pairs for en/decoding so encode() -> decode() and vice-versa
with open("img.txt","rb") as file:
    content = BytesIO()
    content.write(
        file.read().decode("unicode_escape").encode("latin-1")
    )
    # seek back, so the BytesIO() can return back the full content
    content.seek(0)
    # then simply read as if using a file handle
    with Image.open(content) as img:
        img.show()