Computational photography · Datasets and code
Motion Blur Datasets and MATLAB Code
Datasets and code for evaluating motion deblurring and capture strategies. The archive includes high-speed videos, coded-exposure images, and varying-exposure photographs, captured using a static camera.
Averaging high-speed frames simulates a finite exposure. Summing selected frames simulates coded exposure; shifting frames before averaging can simulate camera motion. Moving resolution-chart footage supports evaluation of reconstruction detail and noise in homogeneous regions. See Optimal Single Image Capture for Motion Deblurring for the analysis.
Exposure codes
and reconstruction
Optimal exposure codes cover lengths up to 100 with 50% on/off patterns. For lengths up to 24, all candidate codes can be tested; larger lengths use a random search over one million codes. The objective maximizes the minimum DFT magnitude of the zero-padded code.
The MATLAB/C code-search implementation searches for an optimal code of a given length.
The coded-exposure deblurring package demonstrates reconstruction when blur exceeds the code length. The minimum resolvable blur increases, but the example method avoids introducing ghosting or deconvolution artifacts.
02
High-speed capture
The original frame archives and their simulation-code downloads are unavailable. The original dataset listing also included an outdoor-car sequence.
03
Coded exposure with a Canon camera
Captured using a Canon camera and a ferro-electric shutter. Select an example to download the complete MATLAB code and the included inputs.
Other
capture setups
The original archive also describes a moving Lena-poster example captured with a Point Grey Dragonfly2 camera in trigger mode 5.
Varying-exposure datasets include a Canon Rebel XT Auto Exposure Bracketing capture of a person standing up and an outdoor truck. Point Grey Flea2 captures include an outdoor car and an indoor toy car with occlusions.
05
Aligned face photographs
Select the images to obtain the original files directly.
MATLAB/C
resources
- Coded-exposure deblurring code and images · SIGGRAPH 2006
- Optimal code search · MATLAB/C
- Optimal code sequences
The original simulation examples show how to synthesize coded-exposure images from high-speed videos and resample the blurred image to the code size before deblurring. The associated high-speed frame downloads are unavailable.
Related research
- Coded exposure for motion deblurring · SIGGRAPH 2006
- Dappled Photography · SIGGRAPH 2007
- Invertible Motion Blur in Video · SIGGRAPH 2009
- Motion deblurring and super-resolution · CVPR 2007
- Optimal Single Image Capture for Motion Deblurring · CVPR 2009
- Optimized codes for PSF estimation and invertibility · CVPR 2009
Software license
Copyright 2006-2010 Mitsubishi Electric Research Laboratories All Rights Reserved.
Permission to use, copy and modify softwares linked from this page and its documentation without fee for educational, research and non-profit purposes, is hereby granted, provided that the above copyright notice and the following three paragraphs appear in all copies.
To request permission to incorporate this software into commercial products contact: Vice President of Marketing and Business Development; Mitsubishi Electric Research Laboratories (MERL), 201 Broadway, Cambridge, MA 02139.
IN NO EVENT SHALL MERL BE LIABLE TO ANY PARTY FOR DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING LOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN IF MERL HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
MERL SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS ON AN "AS IS" BASIS, AND MERL HAS NO OBLIGATIONS TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS OR MODIFICATIONS.