1: \begin{abstract}

2: In recent years inpainting-based compression methods have been shown to 

3: be a viable alternative to classical codecs such as JPEG and JPEG2000.

4: Unlike transform-based codecs, which store coefficients in the transform 

5: domain, inpainting-based approaches store a small subset of the original 

6: image pixels and reconstruct the image from those by using a suitable 

7: inpainting operator.

8: A good candidate for such an inpainting operator is homogeneous diffusion 

9: inpainting, as it is simple, theoretically well-motivated, and can achieve 

10: good reconstruction quality for optimized data. However, a major challenge 

11: has been to design fast solvers for homogeneous diffusion inpainting that 

12: scale to 4K image resolution ($3840 \times 2160$ pixels) and are real-time 

13: capable. We overcome this with a careful adaptation and fusion of two of the 

14: most efficient concept from numerical analysis: multigrid and domain 

15: decomposition. 

16: Our domain decomposition algorithm efficiently utilizes GPU parallelism by 

17: solving inpainting problems on small overlapping blocks.  Unlike simple block 

18: decomposition strategies such as the ones in JPEG, our approach yields block 

19: artifact-free reconstructions. Furthermore, embedding domain decomposition 

20: in a full multigrid scheme provides global interactions and allows us to 

21: achieve optimal convergence by reducing both low- and high-frequency errors 

22: at the same rate.

23: We are able to achieve 4K color image reconstruction at more than $60$ frames 

24: per second even from very sparse data - something which has been previously 

25: unfeasible. 

26: \end{abstract}

27: