简体繁体 English

在HOG特征中重新采样图像需要什么？

[英]What is the need of re-sampling the image in HOG features?

原文 2016-05-26 12:02:04 6 1 image-processing/ machine-learning/ computer-vision

I read Dalal and Triggs paper for HOG description and a blog byChris McCormick regarding the same.我阅读了Dalal 和 Triggs论文中的 HOG 描述以及Chris McCormick 撰写的关于相同内容的博客。 The blog says that the image needs to be re-sampled at different scales to recognize different person.该博客说，图像需要以不同的比例重新采样以识别不同的人。

My question is: Already we have a window which we place on the image having a size of 64*128 and which slides over the image.我的问题是：我们已经有一个窗口，我们将其放置在大小为 64*128 的图像上并在图像上滑动。 Then why re-sampling instead of sliding the whole window over the image which can detect the persons instead.那么为什么要重新采样而不是在可以检测人物的图像上滑动整个窗口。 ? ?

Please rectify if I am wrong, thanks in advance !!如有不对请指正，先谢过！！

1 个解决方案

You're right about the fact that the size of 64*128 is trained to be classified as either 'person' or 'non person'.您对 64*128 的大小经过训练被归类为“人”或“非人”这一事实是正确的。 But do all the persons in real world images always come in a handy 64*128 size?但是，现实世界图像中的所有人都总是使用方便的 64*128 尺寸吗？

That is where the scaling comes to play.这就是缩放发挥作用的地方。 By progressively making image smaller, the same 64*128 pixel region will cover larger area in the original image allowing detection of multiple sizes people.通过逐渐缩小图像，相同的 64*128 像素区域将覆盖原始图像中更大的区域，从而可以检测多种尺寸的人。

For example,Here is an example from one of my models after running the detection on multiple scales.例如，这是我的一个模型在多个尺度上运行检测后的示例。 The result presented is after applying non-maximal supression to weed out extreneous detection windows.呈现的结果是在应用非最大抑制以清除无关紧要的检测窗口之后。