Method and device for generating a super-resolution version of a low resolution input data structure

Patent metadata

Field	Value
Publication number	US-9105086-B2
Application number	US-201313889460-A
Country	US
Kind code	B2
Filing date	May 8, 2013
Priority date	May 10, 2012
Publication date	Aug 11, 2015
Grant date	Aug 11, 2015

Abstract

The invention relates to an improved robustness in upscaling the resolution of regularly sampled multi-dimensional signals, where a single low-resolution signal is available. These methods are referred to as example-based super-resolution or single-image super-resolution. A method for super-resolving a single image comprises three stages. First, an interpolation-based up-scaling of the input image is performed, followed by a Local De-noising step in contour regions. The second stage comprises extrapolation through cross-scale block matching, wherein an extrapolated high-frequency band is obtained that is de-noised through regularization in the same contour regions. The third stage comprises adding the contributions of the low-frequency band of the high-resolution image and the extrapolated high-frequency band.

First claim

The invention claimed is:

1. A method for performing super-resolution processing of a low resolution input data structure of digital 1D, 2D or 3D data, comprising steps of
detecting one or more regions in the low resolution input data structure where aliasing is likely to occur, and generating a de-noising mask that indicates the detected one or more regions in the low resolution input data structure;
splitting the input data structure into a low-frequency input data structure a…

Assignees

• Thomson Licensing

Inventors

• Salvador Jordi
• Kochale Axel
• Borsum Malte

Classifications

• G06T3/4053 Primary

  Physics · mapped topic

• G06T5/003 Primary

  Physics · mapped topic

• G06T5/70

  Physics · mapped topic

• G06T3/4007

  Physics · mapped topic

• G06T5/73 Primary

  Physics · mapped topic