It seems you're about deconvolution. Say, we have convolution g(x)a = h where g is an original image, a - camera aperture and h - sensed ('blurred') image, (x) - convolution operation.
Deconvolution is computing G with known A and H, and can be done in many ways.
One is based on fact that Fg(i) * Fa(i) = Fh(i), where F is fourier transform. Obviously Fg(i) = Fh(i) / Fa(i).
In practice, deconvolution greatly increases noise so there's need in noise suppressing algorithm to use with.
发布评论
评论（1）
看来你是关于反卷积。比如说，我们有卷积 g(x)a = h，其中 g 是原始图像，a - 相机光圈和 h - 感测（“模糊”）图像，(x) - 卷积运算。
反卷积是用已知的A和H计算G，并且可以通过多种方式完成。
一种是基于 Fg(i) * Fa(i) = Fh(i) 的事实，其中 F 是傅立叶变换。显然 Fg(i) = Fh(i) / Fa(i)。
实际上，反卷积会大大增加噪声，因此需要使用噪声抑制算法。
It seems you're about deconvolution. Say, we have convolution g(x)a = h where g is an original image, a - camera aperture and h - sensed ('blurred') image, (x) - convolution operation.
Deconvolution is computing G with known A and H, and can be done in many ways.
One is based on fact that Fg(i) * Fa(i) = Fh(i), where F is fourier transform. Obviously Fg(i) = Fh(i) / Fa(i).
In practice, deconvolution greatly increases noise so there's need in noise suppressing algorithm to use with.