This code implements the pipeline described in the paper "Decomposing Time-Lapse Paintings into Layers" by Jianchao Tan, Marek Dvorožňák, Daniel Sýkora, Yotam Gingold from SIGGRAPH 2015. The pipeline is divided into two stages.
- Input: raw time-lapse video
- Output: albedo video
The substeps are:
- Color shift the whole sequence
- Extract keyframes and color shift each sub-sequence
- For each sub-sequence, perform moving std. deviation and moving median
- Whole sequence L0 smoothing
- Perform albedo conversion
- Input: albedo video
- Output: KM layers and PD layers
The programs are:
- PD layer extraction and KM layer extraction
- PD using the spatial coherency solution: The 3-by-3 layer extraction described in the paper
- OpenCV 2.4
- Eigen 3
- JsonCpp 0.5
- zlib
- Bottleneck:
pip install bottleneck - PIL or Pillow (Python Image Library):
pip install Pillow - NumPy
- LAPACK