harshsinh / sparse-inpainting Goto Github PK

This code is submission for assginment 1 for the course EE698K: Modeling and Representation Techiques for Images.

• ./code/src: contains all of the source files, which are
  • irls.cxx and omp.cxx : Function Implementation for IRLS and OMP
  • sparse_inpainting.cxx : Sparse reconstruction based Inpainting implementation.
  • main.cxx : The main driver code for the program
  • create_mask.cxx : OpenCv based program that lets you create masks on custom images.
• ./code/include: contains the corresponding header files.
• ./tools/ : contains various python scripts.
  • makegrey.py : Take images stored in ./images/color/ and creates their corresponding grayscale images and stores them in ./images/gray
  • samplepatch.py : Takes the images stored in ./images/gray, randomly samples 256 different 8x8 patches and stores them in ./images/dictionary/ in a numbered fashion. It also creats an image ./images/dictionary/dictionary.png whoes columns are the 8x8 patches vectorized.
  • makemosaic.py : Takes the images produced from the previous script and generates ./images/dictionary/dictionary128.png and ./images/dictionary/dictionary256.png which have 16x8 and 16x16, 8x8 original patches as their elements.
  • assessquality.py : Runs the NIQE assessment on the images inside the ./images/results folder and displays the results on terminal.

• Requirements

  • python, g++
  • OpenCV
  • skimage, numpy, scipy

• To Compile
  From the root folder run the following commands

  • mkdir build
  • cd build
  • cmake ..
  • make

• To Create Mask

  • ./create_mask PATH_TO_IMAGE
    In PATH_TO_IMAGE specify the locaiton of source image. Draw using a continuous line.
    As of now the limitation on this part is that the mask being drawn has to begin and end close by on the image. This is dut to the fact that I have stored the mask as contrours first.

• To Run Sparse Inpaint

  • ./sparse_inpaint METHOD VALUE DICT_SIZE
    METHOD = IRLS or OMP
    VALUE = epsilon for IRLS and spasity for OMP
    DICT_SIZE = 128 or 256
    This will start displaying three images, which are: the current selected patch, the proposed edited patch and the state of the modified image at present. This will also store the results in a directory ./images/results/