Journal
IEEE SIGNAL PROCESSING LETTERS
Volume 30, Issue -, Pages 588-592Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/LSP.2023.3276645
Keywords
Feature extraction; Transformers; Kernel; Convolution; Image quality; Finite element analysis; Visual systems; Human visual system (HVS); large kernel attention; multiscale feature extraction; NR-IQA
Categories
Ask authors/readers for more resources
No-Reference Image Quality Assessment aims to evaluate image perceptual quality based on human perception. Many studies have used Transformers to simulate the human visual system by assigning different self-attention mechanisms to distinguish image regions. However, the quadratic computational complexity of self-attention is time-consuming and expensive. We propose a lightweight attention mechanism using decomposed large-kernel convolutions to extract multiscale features, and a novel feature enhancement module to simulate the human visual system. Additionally, we compensate for information loss caused by image resizing with supplementary features from natural scene statistics. Experimental results on five standard datasets demonstrate that our proposed method outperforms existing approaches while significantly reducing computational costs.
No-Reference Image Quality Assessment aims to evaluate the perceptual quality of an image, according to human perception. Many recent studies use Transformers to assign different self-attention mechanisms to distinguish regions of an image, simulating the perception of the human visual system (HVS). However, the quadratic computational complexity caused by the self-attention mechanism is time-consuming and expensive. Meanwhile, the image resizing in the feature extraction stage loses the full-size image quality. To address these issues, we propose a lightweight attention mechanism using decomposed large-kernel convolutions to extract multiscale features, and a novel feature enhancement module to simulate HVS. We also propose to compensate the information loss caused by image resizing, with supplementary features from natural scene statistics. Experimental results on five standard datasets show that the proposed method surpasses the SOTA, while significantly reducing the computational costs.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available