Results of the "Boat Video". Downsampling factor r=2.

(a) Low-resolution video -- move cursor over and click to see	(b) TF Model [1]	(c) MAP Model [3]
(d) SDR Model [2]	(e) Our Nuclear Model [5]	(f) True high-resolution image
Comparison of different algorithms on ``Boat'' image with upsampling factor r=2. (a) The low-resolution video. (b) Result of the TF model. (c) Result of the MAP model. (d) Result of the SDR model. (e) Result of our nuclear model. (f) True high-resolution image.

Zoomed in Results of the "Boat Video". Downsampling factor r=2.

(a) Reference low-resolution image	(b) TF Model [1]	(c) MAP Model [3]
(d) SDR Model [2]	(e) Our Nuclear Model [5]	(f) True high-resolution image
Zoomed in comparison of different algorithms on ``Boat'' image with upsampling factor r=2. (a) The reference low-resolution image. (b) Result of the TF model. (c) Result of the MAP model. (d) Result of the SDR model. (e) Result of our nuclear model. (f) True high-resolution image.

Results of the "Boat Video". Downsampling factor r=4.

(a) Low-resolution video -- move cursor over and click to see	(b) TF Model [1]	(c) MAP Model [3]
(d) SDR Model [2]	(e) Our Nuclear Model [5]	(f) True high-resolution image
Comparison of different algorithms on ``Boat'' image with upsampling factor r=4. (a) The low-resolution video. (b) Result of the TF model. (c) Result of the MAP model. (d) Result of the SDR model. (e) Result of our nuclear model. (f) True high-resolution image.

Zoomed in Results of the "Boat Video". Downsampling factor r=4.

(a) Reference low-resolution image	(b) TF Model [1]	(c) MAP Model [3]
(d) SDR Model [2]	(e) Our Nuclear Model [5]	(f) True high-resolution image
Zoomed in Comparison of different algorithms on ``Boat'' image with upsampling factor r=4. (a) The reference low-resolution image. (b) Result of the TF model. (c) Result of the MAP model. (d) Result of the SDR model. (e) Result of our nuclear model. (f) True high-resolution image.

Results of the "Bridge Video". Downsampling factor r=2.

(a) Low-resolution video -- move cursor over and click to see	(b) TF Model [1]	(c) MAP Model [3]
(d) SDR Model [2]	(e) Our Nuclear Model [5]	(f) True high-resolution image
Comparison of different algorithms on ``Bridge'' image with upsampling factor r=2. (a) The low-resolution video. (b) Result of the TF model. (c) Result of the MAP model. (d) Result of the SDR model. (e) Result of our nuclear model. (f) True high-resolution image.

Zoomed in Results of the "Bridge Video". Downsampling factor r=2.

(a) Reference low-reslution image	(b) TF Model [1]	(c) MAP Model [3]
(d) SDR Model [2]	(e) Our Nuclear Model [5]	(f) True high-resolution image
Zoomed in comparison of different algorithms on ``Bridge'' image with upsampling factor r=2. (a) The reference low-resolution image. (b) Result of the TF model. (c) Result of the MAP model. (d) Result of the SDR model. (e) Result of our nuclear model. (f) True high-resolution image.

Results of the "Bridge Video". Downsampling factor r=4.

(a) Low-resolution video -- move cursor over and click to see	(b) TF Model [1]	(c) MAP Model [3]
(d) SDR Model [3]	(e) Our Nuclear Model [5]	(f) True high-resolution image
Comparison of different algorithms on ``Bridge'' image with upsampling factor r=4. (a) The low-resolution video. (b) Result of the TF model. (c) Result of the MAP model. (d) Result of the SDR model. (e) Result of our nuclear model. (f) True high-resolution image.

Zoomed in Results of the "Bridge Video". Downsampling factor r=4.

(a) Reference low-resolution image	(b) TF Model [1]	(c) MAP Model [3]
(d) SDR Model [2]	(e) Our Nuclear Model [5]	(f) True high-resolution image
Zoomed in Comparison of different algorithms on ``Bridge'' image with upsampling factor r=4. (a) The reference low-resolution image. (b) Result of the TF model. (c) Result of the MAP model. (d) Result of the SDR model. (e) Result of our nuclear model. (f) True high-resolution image.

Original "Real Videos". Downloaded from Peyman Milanfar [4]

Results of the "Text Video". Upsampling factor r=2 (top row) and r=4 (second row).

(a) TF Model [1]	(b) MAP Model [3]	(c) SDR Model [2]	(d) Our Nuclear Model [5]
Comparison of different algorithms on "Text Video". Top row with upsampling factor r=2 and second row with upsampling factor r=4. (a) The TF model, (b) The MAP model, (c) The SDR model and (d) Our nuclear model.

Results of the "Disk Video". Upsampling factor r=2 (top row) and r=4 (second row).

(a) TF Model [1]	(b) MAP Model [3]	(c) SDR Model [2]	(d) Our Nuclear Model [5]
Comparison of different algorithms on "Text Video". Top row with upsampling factor r=2 and second row with upsampling factor r=4. (a) The TF model, (b) The MAP model, (c) The SDR model and (d) Our nuclear model.

Results of the "Alpaca Video". Upsampling factor r=2 (top row) and r=4 (second row).

(a) TF Model [1]	(b) MAP Model [3]	(c) SDR Model [2]	(d) Our Nuclear Model [5]
Comparison of different algorithms on "Alpaca Video". Top row with upsampling factor r=2 and second row with upsampling factor r=4. (a) The TF model, (b) The MAP model, (c) The SDR model and (d) Our nuclear model.

Results of the "Eia Video". Upsampling factor r=4.

(a) TF Model [1]	(b) MAP Model [3]
(c) SDR Model [2]	(d) Our Nuclear Model [5]
Comparison of different algorithms on "Eia Video". (a) The TF model, (b) The MAP model, (c) The SDR model and (d) Our nuclear model.

Results of the "Books Video". Upsampling factor r=2.

(a) The low-resolution video -- move cursor over and click to see (Source File MPG )	(b) TF Model [1]	(c) MAP Model [3]
	(d) SDR Model [2]	(e) Our Nuclear Model [5]
The "Books Video" and comparison of different algorithms on the "Books Video". (a) The low-resolution video, (b) The TF model, (c) The MAP model, (d) The SDR model and (e) Our nuclear model.

Zoomed in Results of the "Books Video" - part one. Upsampling factor r=2.

	(b) TF Model [1]	(c) MAP Model [3]
(a) Low-resolution image	(d) SDR Model [2]	(e) Our Nuclear Model [5]
The first zoomed-in comparison of different algorithms on the "Books Video". (a) The low-resolution image, (b) The TF model, (c) The MAP model, (d) The SDR model and (e) Our nuclear model.

Zoomed in Results of the "Books Video" - part two. Upsampling factor r=2.

	(b) TF Model [1]	(c) MAP Model [3]
(a) Low-resolution image	(d) SDR Model [2]	(e) Our Nuclear Model [5]
The second zoomed-in comparison of different algorithms on the "Books Video". (a) The low-resolution image, (b) The TF model, (c) The MAP model, (d) The SDR model and (e) Our nuclear model.

References

• [1] R. H. Chan, S. D. Riemenschneider, L. Shen, and Z. Shen, Tight frame: an efficient way for high-resolution image reconstruction, Applied and Computational Harmonic Analysis, 17 (2004), pp. 91–115.
• [2] Y.-R. Li, D.-Q. Dai, and L. Shen, Multiframe super-resolution reconstruction using sparse directional regularization, Circuits and Systems for Video Technology, IEEE Transactions on, 20 (2010), pp. 945–956.
• [3] Z. Ma, R. Liao, X. Tao, L. Xu, J. Jia, and E. Wu, Handling motion blur in multi-frame super-resolution, in 2015 IEEE Conference on Computer Vision and PatternRecognition (CVPR), June 2015, pp. 5224–5232.
• [4] P. Milanfar http://www.soe.ucsc.edu/~milanfar/software/sr-datasets.html
• [5] R. Zhao, and R. H. Chan, A Nuclear-norm model for multi-frame super-resolution reconstruction from video clips.

Download all the data