Face Super-Resolution via Triple-Attention Feature Fusion Network

Zhao, Kanghui; Lü, Tao; Zhang, Yanduo; Wang, Yu; Wang, Yuanzhi

doi:10.1587/transfun.2021eal2056

Cited by 2 publications

(2 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the purpose of verifying that the reconstruction of our method is better, we compare the super-resolution (SR) images generated by our method with those generated using other algorithms, including Bicubic, RCAN [7], HM-RFN [12], SPARNet [8], EDGAN [10], TAFFN [9]. These networks are trained under the same condition (dataset, optimizer, training time, etc.…”

Section: Compared With State-of-the-artsmentioning

confidence: 99%

“…It enables the convolutional layer to adaptively guide features related to key face structures while paying less attention to those regions with fewer features. Zhao et al [9] introduced a triple-attention mechanism to guide face SR reconstruction by fusing different attention feature maps. Yang et al [10] constructed a discriminative enhanced generative adversarial network (EDGAN) to reconstruct HR face images through the mechanism of adversarial learning.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Face Hallucination via Multi-Scale Structure Prior Learning

YAO

Lü

Zhao

et al. 2023

IEICE Trans. Fundamentals

Self Cite

View full text Add to dashboard Cite

Recently, the face hallucination method based on deep learning understands the mapping between low-resolution (LR) and highresolution (HR) facial patterns by exploring the priors of facial structure. However, how to maintain the face structure consistency after the reconstruction of face images at different scales is still a challenging problem. In this letter, we propose a novel multi-scale structure prior learning (MSPL) for face hallucination. First, we propose a multi-scale structure prior block (MSPB). Considering the loss of high-frequency information in the LR space, we mainly process the input image in three different scale ascending dimensional spaces, and map the image to the high dimensional space to extract multi-scale structural prior information. Then the size of feature maps is recovered by downsampling, and finally the multi-scale information is fused to restore the feature channels. On this basis, we propose a local detail attention module (LDAM) to focus on the local texture information of faces. We conduct extensive face hallucination reconstruction experiments on a public face dataset (LFW) to verify the effectiveness of our method.

show abstract