适用于照片而非扫描图像的 OCR 库
有谁知道可以处理彩色照片（而不是扫描页面）的 OCR 库？
在我看来，大多数图书馆都处理黑白图像，并期望它们来自扫描仪。我需要一些可以拍摄广告牌等彩色照片并从中提取文本的东西。
我目前正在考虑将这些图像转换为黑白 tiff，并将它们提供给这个问题<中提到的库之一< /a>，但我对此表示怀疑......这些库可能是围绕图像是扫描文档或扫描书页的假设而设计的。我需要一些可以处理任何方向、不同大小、奇怪的对齐方式等文本的东西。基本上可以从这些图像中提取最多文本的东西 - 它不必 100% 准确。
有什么想法吗？
Does anyone know of an OCR library that can handle colored photos (as opposed to scanned pages)?
It seems to me that most libraries out there work on B&W images and expect them to come from a scanner. I need something that can take a colored photo of, say, a billboard, and extract text from it.
I'm currently considering converting these images to B&W tiffs and feeding them to one of the libraries mentioned in this question, but I'm skeptical... These library are probably designed around the assumption that images are of scanned documents, or scanned book pages. I need something that would work on text in any orientation, varying sizes, weird alignments, etc. basically something that can extract the most text out of these images - it doesn't have to be 100% accurate.
Any ideas?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
您可以使用 unpaper 进行对齐。 gocr 用于 OCR。通常，获取颜色输入的软件会在处理之前将其转换为灰度 - 这只是人类或软件本身执行的预处理步骤。
You can use unpaper for alignment. gocr for OCR. Usually software that gets colour input will convert it to grey scale before processing - it is just a pre-processing step which humans do or the software does itself.