Light Field Image Compression Based on Convolutional Neural Networks and Linear Approximation

Bakir, Nader; Hamidouche, Wassim; Déforges, Olivier; Samrouth, Khouloud; Khalil, Mohamad

doi:10.1109/icip.2018.8451597

Cited by 36 publications

(25 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Among the different view synthesis techniques proposed, learning-based techniques, mainly using a low-rank approximation alone (as in [250]) or combined with CNN-based techniques (as in [260]), have shown to achieve the best coding performance with significant gains when compared to DIBRbased [158], [159] and transform-assisted [253] synthesis. Nevertheless, the DIBR-based solution proposed in [248] has shown competitive RD performance at high bit rates for coding lenslet LFs by making use of a sparse prediction scheme for coding residuals in LF enhancement layers.…”

Section: ) Discussionmentioning

confidence: 99%

“…With respect to other LF coding solutions in the literature (outside the view synthesis category), view synthesisbased solutions have generally shown better RD performance, mainly at low bit rates, compared to PVS-based LF coding solutions. Moreover, the most recent DIBR-based [248] and learning-based solutions [250], [260] have shown to significantly outperform the MuLE solution based on 4D DCT [125] for coding all kinds of LF content, including lenslet LFs.…”

Section: ) Discussionmentioning

confidence: 99%

“…From these results, the proposed solution is shown to achieve 14.8% bit savings compared to PVS-based HEVC, 8.1% compare to the PVS-based solution winner of ICME 2016 grand challenge in [157], and 4.9% compared to the CNN-based solution in [254]. In [260], a lenslet LF coding scheme is proposed using a hybrid scheme for view synthesis. In this case, two set of key subaperture images are uniformly selected from the LF data.…”

Section: ) Learning-based View Synthesismentioning

confidence: 95%

See 2 more Smart Citations

Dense Light Field Coding: A Survey

2020

View full text Add to dashboard Cite

Light Field (LF) imaging is a promising solution for providing more immersive and closer to reality multimedia experiences to end-users with unprecedented creative freedom and flexibility for applications in different areas, such as virtual and augmented reality. Due to the recent technological advances in optics, sensor manufacturing and available transmission bandwidth, as well as the investment of many tech giants in this area, it is expected that soon many LF transmission systems will be available to both consumers and professionals. Recognizing this, novel standardization initiatives have recently emerged in both the Joint Photographic Experts Group (JPEG) and the Moving Picture Experts Group (MPEG), triggering the discussion on the deployment of LF coding solutions to efficiently handle the massive amount of data involved in such systems. Since then, the topic of LF content coding has become a booming research area, attracting the attention of many researchers worldwide. In this context, this paper provides a comprehensive survey of the most relevant LF coding solutions proposed in the literature, focusing on angularly dense LFs. Special attention is placed on a thorough description of the different LF coding methods and on the main concepts related to this relevant area. Moreover, comprehensive insights are presented into open research challenges and future research directions for LF coding. INDEX TERMS Camera array, image compression, light field, plenoptic, video compression.

show abstract

Section: ) Discussionmentioning

confidence: 99%

Section: ) Discussionmentioning

confidence: 99%

Section: ) Learning-based View Synthesismentioning

confidence: 95%

See 1 more Smart Citation

Dense Light Field Coding: A Survey

2020

View full text Add to dashboard Cite

show abstract

“…The three-dimensional volume is then encoded using the 3D DCT scheme on 8 × 8 × 8 blocks, similarly as in the JPEG coding system. Besides conventional coding methods, also an alternative approach [3] exists that uses deep learning to estimate the 2D view from the sparse sets of 4D views. Another approach [4] proposes own sparse coding scheme for the entire 4D LF based on several optimized key views.…”

Section: Related Workmentioning

confidence: 99%

Evaluation of 4D Light Field Compression Methods

Barina¹,

Chlubna²,

Šolony³

et al. 2019

Computer Science Research Notes

View full text Add to dashboard Cite

Light field data records the amount of light at multiple points in space, captured e.g. by an array of cameras or by a light-field camera that uses microlenses. Since the storage and transmission requirements for such data are tremendous, compression techniques for light fields are gaining momentum in recent years. Although plenty of efficient compression formats do exist for still and moving images, only a little research on the impact of these methods on light field imagery is performed. In this paper, we evaluate the impact of state-of-the-art image and video compression methods on quality of images rendered from light field data. The methods include recent video compression standards, especially AV1 and XVC finalised in 2018. To fully exploit the potential of common image compression methods on four-dimensional light field imagery, we have extended these methods into three and four dimensions. In this paper, we show that the four-dimensional light field data can be compressed much more than independent still images while maintaining the same visual quality of a perceived picture. We gradually compare the compression performance of all image and video compression methods, and eventually answer the question, "What is the best compression method for light field data?".

show abstract

“…Zhao et al [18] proposed linear approximation prior, where the non-encoded views are approximated with a weighted sum of coded views. Recently, convolutional neural networks (CNN)-based approaches have been adopted to synthesis the non-coded views [19][20][21], allowing to recover the whole LF views.…”

Section: Introductionmentioning

confidence: 99%

Subjective Evaluation of Light Field Image Compression Methods based on View Synthesis

Bakir

Fezza

Hamidouche

et al. 2019

2019 27th European Signal Processing Conference (EUSIPCO)

Self Cite

View full text Add to dashboard Cite

Light field (LF) images provide rich visual information enabling amazing applications, from post-capture image processing to immersive applications. However, this rich information requires significant storage and bandwidth capabilities thus urgently raises the question of their compression. Many studies have investigated the compression of LF images using both spatial and angular redundancies existing in the LF images. Recently, interesting LF compression approaches based on view synthesis technique have been proposed. In these approaches, only sparse samples of LF views are encoded and transmitted, while the other views are synthesized at decoder side. Different techniques have been proposed to synthesize the dropped views. In this paper, we describe subjective quality evaluation of two recent compression methods based on view synthesis and comparing them to two pseudo-video sequence based coding approaches. Results show that view synthesis based approaches provide higher visual quality than the naive LF coding approaches. In addition, the database as well as subjective scores are publicly available to help designing new objective metrics or can be used as a benchmark for future development of LF coding methods. 1

show abstract

Light Field Image Compression Based on Convolutional Neural Networks and Linear Approximation

Cited by 36 publications

References 16 publications

Dense Light Field Coding: A Survey

Dense Light Field Coding: A Survey

Evaluation of 4D Light Field Compression Methods

Subjective Evaluation of Light Field Image Compression Methods based on View Synthesis

Contact Info

Product

Resources

About