Reconstruction of Digital Document Images from Scribbled Sketches

The presence of artifacts and defects in digital images compromise their
homogeneity. This problem however can be unraveled by having the defective
regions to be recognized manually, removed and filled. Comes image inpainting, this
predicament is no longer an intricate task. Still, the need for human intervention in
identifying defective regions curtails its autonomy and robusity. Thus, the author
intends to discuss some fundamental ideas behind image inpainting, removal and an
ingenious way of filling those blank holes dubbed "texture synthesis". This report
concisely reviews the concept of image inpainting, right from the semi-automatic
removal of defective regions by simulating Photoshop's Magic Wand tool using
color segmentation and filling-in the blank holes as the results of removing non
textual objects with a plausible result to the human eye. This project focuses on
reconstructing digital images with textual components. To achieve this, the author
implements software prototyping in constructing the software program to realize the
objectives using MATLAB and C++. Finally, the author conducted various analyses
on the accuracy and performance ofthe software program.