* "Images" folder contains 28 images, two for each of the videos listed in Table 1 of the paper.


Set1 - Naming format : VideoName_Mean_Images.jpg

	* Each image is a collage of the following 9 images
		- (1) Mean of the input Video
		- (2) A Ground Truth image showing the undistorted static background 
		- Outputs of restoration methods (3)DL (5)LWB (5)SBR-RPCA.   
		- Outputs of the methods we propose (6)CS (7)PEOF (8)CS+PEOF
		- We also tried supplementing our "CS+PEOF" methods with an RPCA stage. The restoration result is shown in (9)CS+PEOF+RPCA.
		  As mentioned in the paper, we could remove the spatio-temporal noisy artifacts from the video but the visual and numerical quality of the mean image improved only marginally.

Set2 - Naming format : VideoName_SSIM_Dissimilarity
		-Intention of this set is to highlight the geometrical distortions introduced by each method.
		- Each individual image in the collage, except the ground truth is created in the following manner
			- 0.7*RestoredImage + 0.3*(1-SSIM_Map).*Red_Color. Hence, wherever the ssim values are lower, you will those regions getting highlighted with brighter red color.
			- You can observe that the current state of the art methods do not reconstruct the background correctly. Particularly, SBR-RPCA gives a very stable video, but is significantly distorted wrt the 	ground truth.



*Please note: If the files are sorted alphabetically or chronologically, VideoName_SSIM_Dissimilarity and  VideoName_Mean_Images will get arranged back to back for each video. It will be easy to toggle between the restoration result and corresponding SSIM dissimilarity map in the image viwer.

Due to storage space limitations, we have included the results only for few sample video sequences. Results for the complete set of video sequences can be accessed at [1], [2].
