Edge-preserving light-field image super-resolution via feature affine transformation network

文献类型： 外文期刊

第一作者： Zou, Rong

作者： Zou, Rong;Dai, Wenjie;Bai, Shenghe;Mu, Senlin

作者机构：

关键词： Light-field image; Super-resolution; Macro-pixel image; Edge map; Feature affine transformation

期刊名称：OPTICS AND LASERS IN ENGINEERING （ 影响因子：4.6； 五年影响因子：4.9 ）

ISSN： 0143-8166

年卷期： 2024 年 172 卷

页码：

收录情况： SCI

摘要： In the practical application of light-field (LF) images, researchers are often plagued by the problem of low spatial resolution. In this study, we propose a novel method for super-resolution of LF images that uses both the macro pixel image (MacPI) and the sub-aperture image (SAI) of the LF. The MacPI is considered to contain rich angular information that can be used to enhance the super-resolution effect. In addition, in order to maintain the edge features of the image, we introduce the edge map into the condition network to generate prior conditions together with the MacPI. Specifically, we design a feature affine-transformation module that uses the conditions as input to obtain affine-transformation parameters, which guide the features extracted from the SAIs in the backbone network. Additionally, the network uses the features of the central SAI combined with the affinetransformed backbone network features to enhance the super-resolution effect. Finally, the super-resolution LF image is recovered using an up-sampling network. Experimental results on synthetic and real-world datasets demonstrate that the proposed method is highly competitive with other mainstream models in terms of both visual and quantitative evaluation.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序