Statistics of Patch Offsets for Image Completion

By Kaiming He. 2012

210 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

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

In Computer Vision and Pattern Recognition

Citations in the corpus

Semantic Image Inpainting with Deep Generative Models (2016, background): “Prior knowledge, such as statistics of patch offsets [11], planarity [13] or low rank (LR) [12] can greatly improve the result as well.”
Combining Markov Random Fields and Convolutional Neural Networks for Image Synthesis (2016, background): “In this case, structural statistics such as [11] may be used for regularizing the synthesized image.”
Palette: Image-to-Image Diffusion Models (2021, background): “Early inpainting approaches [Barnes et al. 2009; Bertalmio et al. 2000; Hays and Efros 2007; He and Sun 2012] work well on textured regions but often fall short in generating semantically consistent structure.”
SinGAN: Learning a Generative Model From a Single Natural Image (2019, background): “Classical examples include denoising [65], deblurring [39], super resolution [18], dehazing [2, 15], and image editing [37, 21, 9, 11, 50].”
Image Inpainting via Generative Multi-column Convolutional Neural Networks (2018, methodology): “To preserve structure, patch priority computation speciﬁes the patch ﬁlling order [4, 7, 8, 21].”
Missing Information Reconstruction of Remote Sensing Data: A Technical Review (2015, methodology): “The patch or pixel locations are optimized in MRF by the use of belief propagation [21] or graph cuts [33].”
Contextual Residual Aggregation for Ultra High-Resolution Image Inpainting (2020, background): “Another stream is relying on texture synthesis techniques, which fills the hole by both extending and borrowing textures from surrounding regions [6, 9, 13, 29, 31].”
Learning Oracle Attention for High-Fidelity Face Completion (2020, background): “Methods in the first category attempt to copy similar patches from the unmasked area to fill in the missing area [37, 20, 11, 5].”
Image completion using planar structure guidance (2014, background): “Replacing or ﬁlling regions in images with plausibly synthesized content is a common image editing operation.”
Robust Image Completion via Deep Feature Transformations (2019, background): “Prior knowledge, such as statistics of patch offsets [61], planarity [62] or low rank (LR) [61] can greatly improve the result as well.”
Video Inpainting of Complex Scenes (2014, methodology): “This idea was subsequently followed by a series of image inpainting methods [22, 25, 29, 36] which use optimisation techniques such as graph cuts algorithms [9], to minimise a global patch-based functional.”
Learning to Incorporate Structure Knowledge for Image Inpainting (2020, background): “2000; Levin, Zomet, and Weiss 2003; Roth and Black 2005) or fill holes by searching and fusing similar patches from the same image or external image databases (Hays and Efros 2008; Barnes et al. 2009; Pritch, Kav-Venaki, and Peleg 2009; He and Sun 2012).”
Neural Architecture Search for Deep Image Prior (2020, background): “Several in-painting works exploit MRF formulations of this kind [22, 18, 24], including methods that source patches from the input [10] or multiple external [17, 14] images, or use random propagation of a few good matched patches [3].”
Globally and locally consistent image completion (2017, background): “…connect curves across holes [Jia and Tang 2003]; exploiting structure-based priority for patch ordering [Criminisi et al. 2004], tile-based search space constraints [Kopf et al. 2012], statistics of patch offsets [He and Sun 2012], and regularity in perspective planar surfaces [Huang et al. 2014].”
The Synthesizability of Texture Examples (2014, background): “nique of dominant offsets proposed in [12] – only the dominant offsets (60 in our implementation) were considered as the labels.”
Image Inpainting : Overview and Recent Advances (2014, methodology): “The method in [39], based on statistics of patch offsets, also gives very good results in this case.”