| Paper (CVPR - spotlight) | Paper (TIP) | Project Page + Video |
Official implementation of "Noise Robust Depth from Focus using a Ring Difference Filter" (CVPR 2017) and "Ring Difference Filter for Fast and Noise Robust Depth From Focus " (TIP 2019).
Note: This code was used for the quantitative analysis in the paper (bokeh detection, unreliable depth rejection, and tree propagation are not included)
Usage:
demo.m
for an example run on the Buddha
setuse_rdf_agg
to false
for CVPR version or to true
[rdf_labels, im_all_in_focus] = RDF_DFF(pathval, regexp, i_start, numimg,
show_fig, resize, use_parallel, use_rdf_agg, use_fast_jwmf, saveimg)
% Inputs:
% pathval: (str) path to input image set
% regexp: (str) format of image naming
% i_start: (int) path to input image set
% numimg: (int) number of images
% show_fig: (bool) [default: true] switch to render intermediate figures
% resize: (float) [default: 0.5] factor to resize image for aligning images
% use_parallel: (bool) [default: true] switch to use parallel processing (need toolbox)
% use_rdf_agg: (bool) [default: false] switch to use RDF-based aggregation
% use_fast_jwmf: (bool) [default: false] switch to use fast weighted median filtering for end
% saveimg: (bool) [default: false] switch to save intermediate images during alignment
%
% Outputs:
% rdf_labels: (matrix of double)
% im_all_in_focus: (image in double)