blocks|key|31167|text|我曾广泛使用mahotas，但目前已切换到scikit-image。我发现scikit-image在功能上比mahotas更广泛(我需要经常测试各种图像处理算法)。Mahotas的主要优点是代码简单，文档良好，这与它的mahotas-principles是一致的。因此，如果您需要基本的功能和更快的操作，请使用Mahotas。我还没有对scipy-nd+image做过详细的研究，但是看看可用的函数列表，skimage会占据上风。|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|31168|entityMap|0|LINK|mutability|MUTABLE|url|http://mahotas.readthedocs.org/en/latest/principles.html#principles-of-mahotas/^0|31|I|0|0^^$0|@$1|2|3|4|5|6|7|L|8|@]|9|@$A|M|B|N|1|O]]|C|$]]|$1|D|3|-4|5|6|7|P|8|@]|9|@]|C|$]]]|E|$F|$5|G|H|I|C|$J|K]]]]

I have used mahotas extensively, but currently switched to scikit-image. I found scikit-image a bit more extensive in its functionality than mahotas (I need to test various image processing algoorithms frequently). Mahotas main advantage is simple codes with good documentation, which is is agreement with its <a href="http://mahotas.readthedocs.org/en/latest/principles.html#principles-of-mahotas/" rel="nofollow" title="mahotas-principles">mahotas-principles</a>. So, if you require basic functions and faster operation, go for Mahotas. I haven't done a detailed study of scipy-nd image, but looking at the list of available functions, skimage would have the upperhand.

blocks|key|4538733|text|我在Python上帮助维护了一个页面，比较了大约20个常见的图像处理库的速度和内存使用情况，其中包括相当多的libvips系统。|type|unstyled|depth|inlineStyleRanges|entityRanges|offset|length|data|4538734|https://github.com/libvips/libvips/wiki/Speed-and-memory-use|4538735|基准非常简单:加载一个5,000+x+5,000像素的RGB+TIFF，从每个边缘裁剪100个像素，收缩10%25，使用3x3卷积进行锐化，然后再次保存。这是一个愚蠢的测试，但实现起来很简单，而且它确实执行了卷积、加载/保存、重采样和像素操作。|4538736|测试是在libvips+wiki上进行的，所以libvips+(和pyvips，Python+binding)当然会胜出，但根据我的经验，结果是一个合理的指导，至少对于像这样的简单任务来说是这样。这对compare+the+code来说也很有趣。|4538737|System++++++++++++++++++++%7C+Run+time+(secs)+%7C+Peak+mem+use+(MB)
---------------------------------------------------------------
libvips+C+8.8+++++++++++++%7C+0.15++++++++++++%7C++40+
pyvips+2.1.6++++++++++++++%7C+0.18++++++++++++%7C++49+
Pillow-SIMD+5.3+++++++++++%7C+0.36++++++++++++%7C+230+
NetPBM+10.0-15.3++++++++++%7C+0.60++++++++++++%7C++75+
sips+10.4.4+++++++++++++++%7C+0.70++++++++++++%7C+268+
GraphicsMagick+1.3.28+++++%7C+0.64++++++++++++%7C+493+
ImageMagick+6.9.7-4+++++++%7C+0.82++++++++++++%7C+463+
OpenCV+3.2++++++++++++++++%7C+0.93++++++++++++%7C+222+
ImageMagick+7.0.8+++++++++%7C+1.37++++++++++++%7C+733+
ImageJ+1.51+++++++++++++++%7C+2.84++++++++++++%7C+770+
scipy+1.2+%2B+Pillow++++++++%7C+4.33++++++++++++%7C+361+|code-block|syntax|javascript|4538738|下面是相同的图形：|4538739|​|4538740|📷|atomic|4538741|4538742|entityMap|0|LINK|mutability|MUTABLE|url|https://libvips.github.io/libvips/|1|2|3|https://github.com/libvips/vips-bench|4|IMAGE|IMMUTABLE|imageUrl|https://ask.qcloudimg.com/http-save/yehe-900000/db6fe9802556a9929a67496fc593e7e2.png|imageAlt^0|1I|7|0|0|0|1O|1|0|0|X|6|2|2S|G|3|0|0|0|0|0|1|4|0|0^^$0|@$1|2|3|4|5|6|7|1E|8|@]|9|@$A|1F|B|1G|1|1H]]|C|$]]|$1|D|3|E|5|6|7|1I|8|@]|9|@$A|1J|B|1K|1|1L]]|C|$]]|$1|F|3|G|5|6|7|1M|8|@]|9|@]|C|$]]|$1|H|3|I|5|6|7|1N|8|@]|9|@$A|1O|B|1P|1|1Q]|$A|1R|B|1S|1|1T]]|C|$]]|$1|J|3|K|5|L|7|1U|8|@]|9|@]|C|$M|N]]|$1|O|3|P|5|6|7|1V|8|@]|9|@]|C|$]]|$1|Q|3|R|5|6|7|1W|8|@]|9|@]|C|$]]|$1|S|3|T|5|U|7|1X|8|@]|9|@$A|1Y|B|1Z|1|20]]|C|$]]|$1|V|3|R|5|6|7|21|8|@]|9|@]|C|$]]|$1|W|3|-4|5|6|7|22|8|@]|9|@]|C|$]]]|X|$Y|$5|Z|10|11|C|$12|13]]|14|$5|Z|10|11|C|$12|E]]|15|$5|Z|10|11|C|$12|13]]|16|$5|Z|10|11|C|$12|17]]|18|$5|19|10|1A|C|$1B|1C|1D|-4]]]]

I help maintain a page on the <a href="https://libvips.github.io/libvips/" rel="nofollow noreferrer">libvips</a> wiki comparing 20 or so common image processing libraries for speed and memory use, including quite a few Python systems.

<a href="https://github.com/libvips/libvips/wiki/Speed-and-memory-use" rel="nofollow noreferrer">https://github.com/libvips/libvips/wiki/Speed-and-memory-use</a>

The benchmark is very simple: load a 5,000 x 5,000 pixel RGB TIFF, crop 100 pixels from each edge, shrink 10%, sharpen with a 3x3 convolution, and save again. It's a silly test, but simple to implement and it does exercise convolution, load/save, resample and pixel manipulation.

The test is on the libvips wiki, so of course libvips (and <a href="https://libvips.github.io/libvips/" rel="nofollow noreferrer">pyvips</a>, the Python binding) wins, but the results are a reasonable guide, at least for simple tasks like this, in my experience. It's also fun to <a href="https://github.com/libvips/vips-bench" rel="nofollow noreferrer">compare the code</a>.

<pre><code>System | Run time (secs) | Peak mem use (MB)
---------------------------------------------------------------
libvips C 8.8 | 0.15 | 40 
pyvips 2.1.6 | 0.18 | 49 
Pillow-SIMD 5.3 | 0.36 | 230 
NetPBM 10.0-15.3 | 0.60 | 75 
sips 10.4.4 | 0.70 | 268 
GraphicsMagick 1.3.28 | 0.64 | 493 
ImageMagick 6.9.7-4 | 0.82 | 463 
OpenCV 3.2 | 0.93 | 222 
ImageMagick 7.0.8 | 1.37 | 733 
ImageJ 1.51 | 2.84 | 770 
scipy 1.2 + Pillow | 4.33 | 361 
</code></pre>

Here's the same thing graphically:

<a href="https://i.stack.imgur.com/UON8C.png" rel="nofollow noreferrer"><img src="https://i.stack.imgur.com/UON8C.png" alt="graph speed mem use"></a>

I'm a bit mystified as to why Python has so many imaging packages implemented and have been trying to find a table comparing and contrasting either features, strengths/weaknesses, speed/performance, application areas, etc between the bunch of them. So far I know of PIL, mahotas, scikits.image, and scipy.ndimage. I suppose we could also include the many bindings that are supported by external libraries such as ImageMagic, OpenCV, Cairo, etc but let's just start with stuff written specifically for Python. Does anyone know first hand, or is familiar with a blog post or some central location where all these packages are compared and distinguished from one another.

Comparing and Contrasting Python's many Image Processing Libraries

图像处理

Python

我有点困惑，为什么Python实现了这么多的图像包，并且一直在试图找到一个表格，比较它们之间的功能、优缺点、速度/性能、应用领域等。到目前为止，我知道了PIL、mahotas、scikits.image和scipy.ndimage。我想我们也可以包含许多由外部库支持的绑定，比如ImageMagic、OpenCV、Cai...

问比较和对比Python的多个图像处理库
EN

社区

活动

资源

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问比较和对比Python的多个图像处理库EN

社区

活动

资源

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问比较和对比Python的多个图像处理库
EN