Special issue on deep learning for emerging big multimedia super-resolution

Bellandi, Valerio; Chehri, Abdellah; Cuomo, Salvatore; Jeon, Gwanggil

doi:10.1007/s00530-021-00814-5

Cited by 2 publications

(2 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With the evolution of deep learning (DL) approaches such as neural networks, there has been a lot of growth in image processing applications domains, such as astronomical observation image and video enhancement [11], [12].. The advanced DL methods promise enhanced computational power with the ability to process big data to overcome existing issues in the SR domain.…”

Section: Introductionmentioning

confidence: 99%

“…For fast speed, Dong et al [18] introduced the FSRCNN method by inserting a deconvolution layer in the last stage of the network. Jiu et al [19] proposed dense residual network [12] (RDN) and employ large kernel size deconvolution, which reduces the computational complexity of the network and requires less memory. Further, for SISR, Lan et al [20] introduced cascading residual network (CARN) with the help of local and global cascading, which requires less network parameters and achieves better results.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Comparative Analysis of Deep Convolution Networks Based Image Super-Resolution Techniques

Sindge¹,

Dutta²,

Saini³

2023

Artificial Intelligence and Communication Technologies

View full text Add to dashboard Cite

Single Image Super-Resolution (SISR) is gaining huge attention in the digital age across various application domains such as surveillance, medical imaging, and agriculture. Numerous SR methods based on deep learning were used by existing researchers to improve image resolution. Literature shows that deep convolutional neural networks (CNNs) perform exceptionally well to handle degraded images. In this study, CNN-based methods from a deep learning environment are compared to reconstruct the High-Resolution (HR) images. Observations show that SRCNN and FSRCNN can achieve considerable image quality after reconstruction; however, performance is limited to small datasets due to shallow network parameters. Furthermore, VDSR and LapSRN were also utilized against heavy datasets due to their huge computational efficiency.

show abstract

Section: Introductionmentioning

confidence: 99%