Awesome

ImageProcessing_ReflectionRemoval

Hoding my CA scripts

README _________________________________________________________________________________________ NOTE 1: Pls make sure that all the '.m' scripts are under a same directory. NOTE 2: You can add your own testing images under the path of /test_images; Please classify them into two classes, i.e., images for intrinsic images decomposition purpose (under /test_images/intrinsic_images, and should be 'png’ format) and images for reflection removal purpose (under /test_images/reflection_removal, and should be 'jpg' format). NOTE 3: To use MIT’s Local Mean Square Error (LMSE)[5] method to test the the 'quality' of output images, please refer the ATTACHMENT. RUN THE CODE: 1) Open 'main.m' in MATLAB; 2) Change the value of lambda which control the smoothness if you want; 3) Run 'main.m', wait for 'ALL DONE' shown in the command window; 4) Check the detailed results in 'results_log.txt' & 'results_images' folder. Copyrights info: 1) This code is written by myself and it is adapted from algorithms published in "Single Image Layer Separation using Relative Smoothness" (Y. Li et al., CVPR 2014) and "Ground truth dataset and baseline evaluations for intrinsic image algorithms" (R. Grosse et al., ICCV 2009). 2) Some testing images under "test_images" folder are copied from MIT intrinsic image dataset and from the testing dataset of " Exploiting Reflection Change for Automatic Reflection Removal" (Y. Li et al., ICCV 2013). CHI JI E0001795@u.nus.edu _________________________________________________________________________________________ ATTACHMENT _________________________________________________________________________________________ NOTE: Copied from MIT intrinsic image dataset. The code is under the 'MIT-intrinsic' folder. The data are available at: http://people.csail.mit.edu/rgrosse/intrinsic/intrinsic-data.tar.gz Unpack the tarballs and merge if necessary. The top-level folder, named MIT-intrinsic by default, should contain the README, four python files, the data folder, and two empty results folders. The four python files are: comparison.py: the script for performing hold-one-out cross-validation. intrinsic.py: all of the intrinsic image algorithms, along with functions for reading the data and computing the error scores poisson.py: functions for solving the Poisson equation using least-squares or L1. html.py: a utility for saving results to HTML. After installing the required packages (see below), you should be able to reproduce most of the results from the paper by running comparison.py: cd MIT-intrinsic python comparison.py This will evaluate the algorithms using hold-one-out cross-validation. It prints results to the console, and also saves the shading/reflectance decompositions and their error scores to the HTML file results/index.html. If you set the USE_L1 variable (defined in comparison.py) to True, it will use the L1 penalty for reconstruction rather than least squares. In this case, the outputs will be saved to results_L1/index.html. We have done our best to provide a code base which is readable, compact, and easy to extend. Please send your questions and comments to Roger Grosse (rgrosse@mit.edu).

_________________________________ Installation __________________________________ To run the code, you will need Python as well as the following Python libraries: NumPy SciPy PyPNG PyAMG __________________________________________________________________________________