Work

Training-based Image Upsampling using Kernels

Year: 2019
Type: article
Abstract: This paper proposes an issue of upsampling which populates missing pixel components at the zero-padded images. This process is called image upsampling (or image reconstruction). In this paper, a kernel design approach is studied in detail, which is obtained based on least-squares method by exploiting training set to achieve desired upsampling performance. To meet the complexity requirement, tradeoff approach is considered during the kernel design. Performance of the proposed method is compared with that of other conventional upsampling methods such as nearest neighbor, bilinear interpolation, and bicubic interpolation, in terms of objective and subjective quality metrics. Simulation results inform that the proposed method provides outstanding performance when it is compared with conventional methods.
Language: English
Open Access status: diamond
OpenAlex ID: W6942101593
FWCI:
Cites: 14
Cited by:
Related to: 10