第7课：4周1次

1点到为止，主题内容讲完，再切到convolution in deep network：1.3.2.1 Separable Convolution ，再切回introduction中的inpainting (predictive filtering), intrinsic, dynamic。

inpaint讲完了pipeline，todo

1 Introduction

1.1 Why filtering?

1.1.1 denoising

1. why noisy?
   1. Devices like high-resolution cameras, electron microscopes, and DNA sequencers are capable of producing measurements in the thousands to millions of feature dimensions.
   2. But when these devices are pushed to their limits, taking videos with ultra-fast frame rates at very low-illumination, probing individual molecules with electron microscopes, or sequencing tens of thousands of cells simultaneously, each individual feature can become quite noisy.
   • 万维钢 精英日课 5: 效率是一种脆弱游戏
2. why could the signal be recovered from noisy input?
   1. Speaking loosely, if the “latent dimension” of the space of objects under study is much lower than the dimension of the measurement, it may be possible to implicitly learn that structure, denoise the measurements, and recover the signal without any prior knowledge of the signal or the noise.

      1. spatial domain

         

      2. feature domain, 参见下文的1.1.2 inpainting

   2. 物体是有结构的，噪声是无结构的独立的。

1.1.2 Inpainting

1. 3.3.1 MISF, cvpr22

1.1.3 Intrinsic image decomposition

1.1.4 High Dynamic Range Imaging

1.1.5 Image pyramids

1. Scale-space representation allows coarse-to-fine operations

   

2 Image Smoothing

Split an image into

• large-scale features, structure