Super-resolution for mixed-resolution multiview image plus depth data using a novel two-stage high-frequency extrapolation method for occluded areas

In this paper, we propose a new super-resolution algorithm based on wavelet coefficient. The proposed algorithm uses discrete wavelet transform (DWT) to decompose the input low-resolution image sequences into four subband images, including LL, LH, HL, HH. Then the input images have been processed by the 3DSKR (Three Dimensional Steering Kernel Regression) super resolution (SR) algorithm, and the result replaces the LL subband image, while the three high-frequency subband images have been interpolated. Finally, combining all these images to generate a new high-resolution image by using inverse DWT. Proposed method has been verified on Calendar and Foliage by Matlab software platform. The peak signal-to-noise (PSNR), structural similarity (SSIM) and visual results are compared, and show that the computational complexity of the proposed algorithm decline by 30 percent compared with the existing algorithm to obtain the approximate results.

Download Full-text

A Two-Stage Multi-loss Super-Resolution Network for Arterial Spin Labeling Magnetic Resonance Imaging

Lecture Notes in Computer Science - Medical Image Computing and Computer Assisted Intervention – MICCAI 2019 ◽

10.1007/978-3-030-32248-9_2 ◽

2019 ◽

pp. 12-20 ◽

Cited By ~ 2

Author(s):

Zheng Li ◽

Qingping Liu ◽

Yiran Li ◽

Qiu Ge ◽

Yuanqi Shang ◽

...

Keyword(s):

Magnetic Resonance Imaging ◽

Magnetic Resonance ◽

Arterial Spin Labeling ◽

Super Resolution ◽

Spin Labeling ◽

Resonance Imaging ◽

Two Stage

Download Full-text

Image Super Resolution Reconstruction Based on Improved Invertible Rescaling Net

10.21203/rs.3.rs-545193/v1 ◽

2021 ◽

Author(s):

Taiping Mo ◽

Dehong Chen

Keyword(s):

Feature Extraction ◽

High Frequency ◽

Interpolation Method ◽

Super Resolution ◽

Frequency Information ◽

Ill Posed ◽

Image Edge ◽

Image Super Resolution ◽

Improved Model ◽

Image Details

Abstract The Invertible Rescaling Net (IRN) is modeling image downscaling and upscaling as a unified task to alleviate the ill-posed problem in the super-resolution task. However, the ability of potential variables of the model embedded high-frequency information is general, which affects the performance of the reconstructed image. In order to improve the ability of embedding high-frequency information and further reduce the complexity of the model, the potential variables and feature extraction of key components of IRN are improved. Attention mechanism and dilated convolution are used to improve the feature extraction block, reduce the parameters of feature extraction block, and allocate more attention to the image details. The high frequency sub-band interpolation method of wavelet domain is used to improve the potential variables, process and save the image edge, and enhance the ability of embedding high frequency information. Experimental results show that compared with IRN model, improved model has less complexity and excellent performance.

Download Full-text