什么是“无为”?卷积核
如果我尝试在频率空间中执行卷积核 - 什么是“不执行任何操作”的内核。换句话说,如果我在应用内核并在频率空间中对其进行归一化后查看图像,我只想查看原始傅里叶变换
它是单位矩阵吗?我的内核是 3x3
谢谢
If I am trying to do a convolution kernel in Frequency space - what is the "do-nothing" kernel. In other words, if I view the image after applying the kernel, and normalizing it in Frequency space, I just want to see the raw Fourier transform
Is it the identity matrix? my kernel is 3x3
Thanks
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(2)
一个什么都不做的 3x3 内核将是:
我希望我正确理解了你的问题 - 我不确定为什么你会想要这样的内核,因为完全跳过卷积要容易得多。
A do-nothing 3x3 kernel will be:
I hope I understood your question correctly - I'm not sure why you would want such a kernel, when it's much easier to just skip the convolution entirely.
“无所事事”的卷积核是狄拉克δ函数:“δ(x)”。
分享的解决方案 mark-ransom 就是这样!与狄拉克德尔塔卷积的任何信号都与原始信号相同。这适用于任何 n 维的卷积。
狄拉克 delta 还有许多其他有趣的性质:
另请参阅 如何创建 δ 卷积神经网络层
The "do-nothing" convolution kernel is the Dirac delta function: "δ(x)".
The solution mark-ransom shared is just that! Any signal convolved with the Dirac delta is identical to the original signal. This applies to convolution in any n-dimension.
The Dirac delta has many other interesting properties:
Also see how to create δ convolutional neural network layer