Image super-resolution based on local self-similarity

Suetake, Noriaki; Sakano, Morihiko; Uchino, Eiji

doi:10.1007/s10043-008-0005-0

Cited by 47 publications

(23 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While this offers an example database of a limited size compared to universal databases, as we show later, this example data is considerably more relevant to the input image being enlarged. Suetake et al [2008] also use the input image to compute an example codebook which is later used to estimate the missing highfrequency band, in a framework similar to [Freeman et al 2002].…”

Section: Previous Workmentioning

confidence: 99%

“…The basic upscaling scheme we use is closely related to the framework previously used by Freeman et al [2002] and others [Tappen et al 2004;Suetake et al 2008]. However, we replace most of the components it uses with novel, application-specific, components that we describe here.…”

Section: Upscaling Schemementioning

confidence: 99%

“…As we explain later, this requires the two images to be spectrally compatible which we ensure when designing U and D. The complement high-frequency content in the input image at the matched patch, H 0 (q) = I 0 (q)−L 0 (q), is used to fill-in the missing higher frequency band in the upsampled image by simply pasting it, i.e., 3.1 Local Self-Similarity Freeman et al [2002] use a universal example database of small patches taken from arbitrary natural images. Barnsley [1988] and others [Vrscay 2002;Freeman et al 2000;Ebrahimi and Vrscay 2007;Suetake et al 2008;Glasner et al 2009] rely on self similarities within the image; small patches in natural images tend to repeat themselves within the image and across its scales. This allows to replace the external database with the input image itself, taken at smaller scales, as the source for example patches.…”

Section: Upscaling Schemementioning

confidence: 99%

See 2 more Smart Citations

Image and video upscaling from local self-examples

2011

View full text Add to dashboard Cite

We propose a new high-quality and efficient single-image upscaling technique that extends existing example-based super-resolution frameworks. In our approach we do not rely on an external example database or use the whole input image as a source for example patches. Instead, we follow a local self-similarity assumption on natural images and extract patches from extremely localized regions in the input image. This allows us to reduce considerably the nearest-patch search time without compromising quality in most images. Tests, that we perform and report, show that the local-self similarity assumption holds better for small scaling factors where there are more example patches of greater relevance. We implement these small scalings using dedicated novel non-dyadic filter banks, that we derive based on principles that model the upscaling process. Moreover, the new filters are nearly-biorthogonal and hence produce high-resolution images that are highly consistent with the input image without solving implicit back-projection equations. The local and explicit nature of our algorithm makes it simple, efficient and allows a trivial parallel implementation on a GPU. We demonstrate the new method ability to produce high-quality resolution enhancement, its application to video sequences with no algorithmic modification, and its efficiency to perform real-time enhancement of lowresolution video standard into recent high-definition formats.

show abstract

Section: Previous Workmentioning

confidence: 99%

Section: Upscaling Schemementioning

confidence: 99%

Section: Upscaling Schemementioning

confidence: 99%

See 1 more Smart Citation

Image and video upscaling from local self-examples

2011

View full text Add to dashboard Cite

show abstract

“…This section describes the self-similarity based SR approach [1], [4], [5], which we aim to speed up in this letter. Our target self-similarity SR approach based on the low frequency (LF)-high frequency (HF) domain relies on the assumption that the LF patch correlates highly with the HF patch [2], [4].…”

Section: Self-similarity Based Super Resolutionmentioning

confidence: 99%

“…In particular, self-similarity based SR methods have attracted interest because they do not require largescale external examples [1]- [5].…”

Section: Introductionmentioning

confidence: 99%

Fast Search of a Similar Patch for Self-Similarity Based Image Super Resolution

Yoo

Choi

et al. 2016

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYIn the self-similarity super resolution (SR) approach, similar examples are searched across down-scales in the image pyramid, and the computations of searching similar examples are very heavy. This makes it difficult to work in a real-time way under common software implementation. Therefore, the search process should be further accelerated at an algorithm level. Cauchy-Schwarz inequality has been used previously for fast vector quantization (VQ) encoding. The candidate patches in the search region of SR are analogous to the code-words in the VQ, and CauchySchwarz inequality is exploited to exclude implausible candidate patches early. Consequently, significant acceleration of the similar patch search process is achieved. The proposed method can easily make an optimal trade-off between running speed and visual quality by appropriately configuring the bypass-threshold.

show abstract