###################################################################### # # # README - Usage for corner_detection and inpaint # # # # Daniel Gusenburger July 2020 # # # ###################################################################### To build both programs, simply run make. 1. Corner detection with the structure tensor ./corner [OPTIONS] [input image file (.pgm format)] Options: -------- -s <float>: Sigma/Noise scale Default: 1.0 -r <float>: Rho/Integration scale Default: 2.5 -c <int>: Type of the corner detector 0 -> Rohr 1 -> Tomasi-Kanade 2 -> Förstner-Harris Default: Förstner-Harris (2) -q <float>: Percentile parameter Meaning depends on whether TPPT is enabled or not Default: 0.1 -m <float>: Mask radius in pixels Default: 0 -C: Flag to enable CNMS, uses -m as mask radius -P: Flag to enable TPPT, uses -q as percentile -M: Output the computed mask (in black/white binary) instead of the corner location -d <char>: 'x' to mark corner locations as crosses 'o' to mark corner locations as circles Default: 'o' -o <string>: Filename for output file. Default: "corners.pgm" 2. Inpainting using edge-enhancing diffusion ###################################################################### # START DISCLAIMER # ###################################################################### # # # The mask image used in this version has to be a black white image, # # where the seed points are marked white and the rest is black. # # The other input image has to be the original image. # # Using this mask and the original image, the initial image is # # created as an image where the mask points are initiated by their # # value in the original image and all other values are initiated # # with a grey value of 127. # # # ###################################################################### # END DISCLAIMER # ###################################################################### ./inpaint [OPTIONS] [input image - original (.pgm format)] [mask image (.pgm format)] Options: -------- -s <float>: Sigma/Noise scale Default: 3.0 -l <float>: Lambda/Contrast parameter Default: 1.0 -g <float>: Gamma/Non-negativity parameter Default: 1.0 -a <float>: Alpha/Dissipitavity parameter Default: 0.49 -n <int>: Number of inner/solver iterations Default: 200 -o <string>: Filename of output image Default: "inpaint.pgm" The algorithm stops if the residual error falls below a set threshold of 0.000001 (10^(-6)) As for the other parameters in the original version: Diffusivity: Charbonnier Time-discretisation: semi-implicit Solver: Conjugate Gradients (CG) Rho is always set to 0 since I examined EED not CED.