自编码器

自编码器的输入对自编码器的结果有显著影响，可以通过输入在一定程度上控制输出。
可以 参考论文Deep Image Prior的代码 https://github.com/DmitryUlyanov/deep-image-prior/blob/master/inpainting.ipynb 。这个代码中使用网格矩阵作为输入，具有生成垂直和水平纹理的功能。

自编码器由于下采样和全连接会导致生成的图像清晰度不够。
EdgeConnect: Generative Image Inpainting with Adversarial Edge Learnig 中有这样一段话
One of the first deep learning methods designed for
image inpainting is context encoder [38], which uses an
encoder-decoder architecture. The encoder maps an image
with missing regions to a low-dimensional feature space,
which the decoder uses to construct the output image. How�ever, the recovered regions of the output image often con�tain visual artifacts and exhibit blurriness due to the in�formation bottleneck in the channel-wise fully connected
layer. This was addressed by Iizuka et al. [22] by reduc�ing the number of downsampling layers, and replacing the
channel-wise fully connected layer with a series of dilated
convolution layers [51]. The reduction of downsampling
layers are compensated by using varying dilation factors.