亲爱的学霸们大家好 这是由Károly Zsolnai-Fehér带来的两分钟论文
Dear Fellow Scholars, this is Two Minute Papers with Károly Zsolnai-Fehér.
本系列我们目前已有130多集
We’re currently more than 130 episodes into the series,
但从没一起讨论过这种算法
and we still haven’t talked about this algorithm.
没有 PatchMatch 工作要怎么继续？
How could we go on for so long without PatchMatch?
那么我们开始吧！
So, let’s do this right now!
你一定会爱上它的
You’ll love this one.
这项技术可以帮我们对已有照片进行做出令人惊奇的修改
This technique helps us to make absolutely crazy modifications to previously existing
对于迄今为止所有广泛应用的照片处理技术来说
photographs and it is one of the landmark papers for all kinds of photo manipulation
这项技术绝对是里程碑式的
which is still widely used to this day.
细想一下工作流程：我们将此图像作为输入
Consider the following workflow: we have this image as an input.
依照文献样式 为残缺处 绘制屋顶轮廓线或修复图像
Let’s mark the roofline for hole filling, or image inpainting as the literature refers to it.
这样残缺的地方就被完美的填补好了
And the hole is now filled with quite sensible information.
现在 我们要标记一些柱子来重塑建筑
Now we mark some of the pillars to reshape the object.
然后 我们把屋顶向上拉
And then, we pull the roof upward.
出来的图像同之前会变得完全不一样
The output is a completely redesigned version of the input photograph.
哇噢 这简直令人不可思议
Wow, absolutely incredible.
整个过程几乎在实时交互式地发生
And the whole thing happens interactively, almost in real time,
但如果我们考虑自从论文发表以来的硬件改进
but if we consider the hardware improvement since this paper was published, it is safe to say that today it
可以肯定的是 即使在一台普通的电脑上 它也能实时运行
runs in real time even on a mediocre computer.
这项工作 通过在算法中 添加附加提示 完成图像修复的部分
And in this piece of work, the image completion part works by adding additional hints to the algorithm,
例如 给物体上我们希望删去的预期形状做标记
for instance, marking the expected shape of an object that we wish to cut out
并填写新的数据
and have it filled in with new data.
或者 显示我们希望编辑的建筑物的形状 这也能指导该技术
Or, showing the shape of the building that we wish to edit also helps the technique considerably.
这项结果令人吃惊 我记得我第一次见到它们的时候
These results are so stunning, I remember that when I had first seen them,
我反复核对了一遍又一遍 因为我简直不敢相信我的眼睛
I had to recheck over and over again because I could hardly believe my eyes.
这种技术不仅提供了这些高质量的结果
This technique offers not only these high quality results,
而且比它的竞争对手要快得多
but it is considerably quicker than its competitors.
为了完成图像修补 大多数算法都会 查找图像中与被删除的区域相似的区域
To accomplish image inpainting, most algorithms look for regions in the image that are similar
并从中获取一些填充信息
to the one that is being removed, and borrow some information from there for the filling process.
此时 加快进程的关键思想之一就是找到良好的对照信息
Here, one of the key ideas that speed up the process is that when good correspondances are found,
如果我们正在查找另一项信息 就不要重启补丁匹配程序
if we are doing another lookup, we shouldn’t restart the patch matching process,
应该尝试在附近搜索
but we should try to search nearby,
因为在那里 我们最有可能找到有价值的信息
because that’s where we are most likely to find useful information.
你知道它最牛的部分是哪里么？
You know what the best part is?
这仅仅是开始：这项技术没有使用任何现代化机器学习
What you see here is just the start: this technique does not use any of the modern machine
所以 在这个极其强大的深度神经网络的时代
learning techniques, so in the era of these incredibly powerful deep neural networks,
可想而知将来我们能获得怎样的高品质解决方案
I can only imagine the quality of solutions we’ll be able to obtain in the near future.
我们生活在一个如此神奇的年代
We are living amazing times indeed.
感谢您的收看和支持 下期我们不见不散
Thanks for watching and for your generous support, and I’ll see you next time!