如何创建深度缩放文件并在谷歌驱动器中保留一个非常大的图像600k x 600k文件？

问题描述

我正在使用带有以下代码的 deepzoom python API-

import os
import deepzoom
# Specify your source image
SOURCE = "https://drive.google.com/uc?export=view&id=1W5UKlO_wZJLNvqsyC0E3zJ5CVlSDp_PG"
DEST = "D:\\mesh.dzi"

# Create Deep Zoom Image creator with weird parameters
creator = deepzoom.ImageCreator(
tile_size=128,tile_overlap=2,tile_format="png",image_quality=0.8,resize_filter="bicubic",)

# Create Deep Zoom image pyramid from source

creator.create(SOURCE,DEST)

请注意，我使用的是 Windows 10 操作系统和 Jupyter Notebook 来运行代码。但显示以下错误-

---------------------------------------------------------------------------
UnidentifiedImageError                    Traceback (most recent call last)
<ipython-input-14-6b5c22482c62> in <module>
  1 # Create Deep Zoom image pyramid from source
----> 2 creator.create(SOURCE,DEST)

~\anaconda3\lib\site-packages\deepzoomtools-2.0.0a2-py3.8.egg\deepzoom\__init__.py in 
create(self,source,destination)
411     def create(self,destination):
412         """Creates Deep Zoom image from source file and saves it to destination."""
--> 413         self.image = PIL.Image.open(safe_open(source))
414         width,height = self.image.size
415         self.descriptor = DeepZoomImageDescriptor(

~\anaconda3\lib\site-packages\PIL\Image.py in open(fp,mode,formats)
2941     for message in accept_warnings:
2942         warnings.warn(message)
-> 2943     raise UnidentifiedImageError(
2944         "cannot identify image file %r" % (filename if filename else fp)
2945     )

UnidentifiedImageError: cannot identify image file <_io.BytesIO object at 0x000001FE552ED9F0>

解决方法

改为尝试 libvips。 libvips 处理大图像没有问题。

libvips documentation.

,

您可以使用 pyvips 从 URL 中读取图像，如下所示：

#!/usr/bin/python3

import urllib.request
import pyvips

URL = 'http://www.rollthepotato.net/~john/IMG_2420.JPG'

input_file = urllib.request.urlopen(URL) 

# pyvips will use this to fetch bytes from the URL
def read_handler(size):
    return input_file.read(size)

# 'sequential' means stream the image during processing
source = pyvips.SourceCustom()
source.on_read(read_handler)
image = pyvips.Image.new_from_source(source,'',access='sequential')

print('writing mesh.dzi ...')
image.dzsave('mesh')

这将从服务器流式传输图像并生成 DZI，而无需先下载。

不过，您需要弄清楚如何从 gdrive 获取正确的 URL。