An Image Inpainting Technique Based on the Fast Marching Method

By Alexandru Cristian Telea. J. Graphics, GPU, & Game Tools, 2004

1,556 citations recorded by the source corpus.

Abstract Digital inpainting provides a means for reconstruction of small damaged portions of an image. Although the inpainting basics are straightforward, most inpainting techniques published in the literature are complex to understand and implement. We present here a new algorithm for digital inpainting based on the fast marching method for level set applications. Our algorithm is very simple to implement, fast, and produces nearly identical results to more complex, and usually slower, known methods. Source code is available online.

Subfields: Computer Graphics and Computer-Aided Design, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/67d0cb47d14150daff08980efbea9f1267d3a4e5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Graphics and Computer-Aided Design

In Computer Vision and Pattern Recognition

Citations in the corpus

Automatic Registration of RGB-D Scans via Salient Directions (2013, neutral): “Holes themselves are detected by searching for connected components in the initially rendered images, while for inpainting we utilize FMM [20] as it is simple and fast.”
UCTGAN: Diverse Image Inpainting Based on Unsupervised Cross-Space Translation (2020, methodology): “Early image inpainting approaches progressively ﬁll in missing regions by searching and pasting the most similar image patches in the background regions under the hypothesis that the contents to be ﬁlled in come from the background areas [1, 21, 2].”
MultiStencils Fast Marching Methods: A Highly Accurate Solution to the Eikonal Equation on Cartesian Domains (2007, background): “This order information is very important in several applications such as ordered thinning skeletonization [10], [27], robotic path planning [5], and image inpainting [28].”
Image Inpainting : Overview and Recent Advances (2014, methodology): “pde-based diffusiOn fast marchinG diffusiOn [10] (in Opencv): http://docs.”
Synthesizing Obama (2017, methodology): “…since in some poses the neck is occluded, we automatically mask out the clothing in every video frame (by means of simple thresholding in HSV space; the same threshold is used in all results) and in-paint the masked region using [Telea 2004], and the OpenCV [Bradski 2000] implementation.”
Text2Room: Extracting Textured 3D Meshes from 2D Text-to-Image Models (2023, methodology): “Thus, we first inpaint small holes with a classical inpainting algorithm [83] and dilate the remaining holes.”
Color-Guided Depth Recovery From RGB-D Data Using an Adaptive Autoregressive Model (2014, positive): “S. Lee et al. [8] ﬁlled out the missing areas by an image inpainting algorithm [41], and extended the joint bilateral ﬁlter to the joint multilateral ﬁlter to improve depth quality and temporal consistency.”
SnapNet: 3D point cloud semantic labeling with 2D deep segmentation networks (2017, methodology): “First, we apply an enhancement procedure in the image space: inpainting, using [46], to ﬁll the missing parts of the acquisition, and denoising from [47].”
MVSS-Net: Multi-View Multi-Scale Supervised Networks for Image Manipulation Detection (2021, methodology): “We apply regular data augmentation for training, including flipping, blurring, compression and naive manipulations either by cropping and pasting a squared area or using built-in OpenCV inpainting functions [46], [47].”
SonoNet: Real-Time Detection and Localisation of Fetal Standard Scan Planes in Freehand Ultrasound (2016, methodology): “To prevent our algorithm from learning the manual annotations placed on the images by the sonographers rather than from the images themselves, we removed all the annotations using the inpainting algorithm proposed in [33].”
Full-frame video stabilization with motion inpainting (2006, methodology): “The actual scanning and composition in the missing area M is achieved using the Fast Marching Method (FMM) [38] as described by [39] in the context of image inpainting.”
Contrastive Learning for Diverse Disentangled Foreground Generation (2022, methodology): “Traditional methods [2,4,43,12,3] typically rely on low-level assumptions and image statistics, leading to over smoothing and results with limited visual semantics.”
Image Denoising and Inpainting with Deep Neural Networks (2012, background): “The stateof-the-art non-blind inpainting algorithms can perform very well on removing text, doodle, or even very large objects [10] [11] [12].”
Mask-Specific Inpainting with Deep Neural Networks (2014, methodology): “Visually our approach performs better than the approaches by [5], [2], [20], [18], [22] and by [1].”
Globally and locally consistent image completion (2017, methodology): “In particular, we employ the fast marching method [Telea 2004], followed by Poisson image blending [Pérez et al. 2003].”
Image Inpainting for Irregular Holes Using Partial Convolutions (2018, methodology): “Non-learning approaches to image inpainting rely on propagating appearance information from neighboring pixels to the target region using some mechanisms like distance field[3, 1, 30].”
A cloud detection algorithm for satellite imagery based on deep learning (2019, methodology): “A more advanced inpainting method (Telea, 2004) was tested, but did not result in significant benefit.”