Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MobileIQA: Exploiting Mobile-level Diverse Opinion Network For No-Reference Image Quality Assessment Using Knowledge Distillation

Zewen Chen, Sunhan Xu, Yun Zeng, Haochen Guo, Jian Guo, Shuai Liu, Juan Wang, Bing Li, Weiming Hu, Dehua Liu, Hesong Li

TL;DR

The proposed MobileIQA is a novel approach that utilizes lightweight backbones to efficiently assess image quality while preserving image details through high-resolution input, and outperforms novel IQA methods on evaluation metrics and computational efficiency.

Abstract

With the rising demand for high-resolution (HR) images, No-Reference Image Quality Assessment (NR-IQA) gains more attention, as it can ecaluate image quality in real-time on mobile devices and enhance user experience. However, existing NR-IQA methods often resize or crop the HR images into small resolution, which leads to a loss of important details. And most of them are of high computational complexity, which hinders their application on mobile devices due to limited computational resources. To address these challenges, we propose MobileIQA, a novel approach that utilizes lightweight backbones to efficiently assess image quality while preserving image details through high-resolution input. MobileIQA employs the proposed multi-view attention learning (MAL) module to capture diverse opinions, simulating subjective opinions provided by different annotators during the dataset annotation process. The model uses a teacher model to guide the learning of a student model through knowledge distillation. This method significantly reduces computational complexity while maintaining high performance. Experiments demonstrate that MobileIQA outperforms novel IQA methods on evaluation metrics and computational efficiency. The code is available at https://github.com/chencn2020/MobileIQA.

MobileIQA: Exploiting Mobile-level Diverse Opinion Network For No-Reference Image Quality Assessment Using Knowledge Distillation

TL;DR

The proposed MobileIQA is a novel approach that utilizes lightweight backbones to efficiently assess image quality while preserving image details through high-resolution input, and outperforms novel IQA methods on evaluation metrics and computational efficiency.

Abstract

With the rising demand for high-resolution (HR) images, No-Reference Image Quality Assessment (NR-IQA) gains more attention, as it can ecaluate image quality in real-time on mobile devices and enhance user experience. However, existing NR-IQA methods often resize or crop the HR images into small resolution, which leads to a loss of important details. And most of them are of high computational complexity, which hinders their application on mobile devices due to limited computational resources. To address these challenges, we propose MobileIQA, a novel approach that utilizes lightweight backbones to efficiently assess image quality while preserving image details through high-resolution input. MobileIQA employs the proposed multi-view attention learning (MAL) module to capture diverse opinions, simulating subjective opinions provided by different annotators during the dataset annotation process. The model uses a teacher model to guide the learning of a student model through knowledge distillation. This method significantly reduces computational complexity while maintaining high performance. Experiments demonstrate that MobileIQA outperforms novel IQA methods on evaluation metrics and computational efficiency. The code is available at https://github.com/chencn2020/MobileIQA.
Paper Structure (20 sections, 2 equations, 6 figures, 7 tables)

This paper contains 20 sections, 2 equations, 6 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Proposed Method
  4. Model Design
  5. (A) Multi-level Feature Perception.
  6. (B) Local Distortion Aware Module.
  7. (C) Multi-view Attention Learning Module.
  8. (D) Image Quality Score Regression.
  9. Knowledge Distillation
  10. Experiments
  11. Datasets
  12. Evaluation Metrics
  13. Implementation Details
  14. Comparisons With State-of-the-Arts
  15. Discussion about the Number of the MAL
  16. ...and 5 more sections

Figures (6)

  • Figure 1: Comparison among SOTA IQA methods on UHD-IQA hosu2024uhd validation set in terms of KROCC, SROCC, PLCC and MACs.
  • Figure 2: Framework of the teacher model (MobileViT-IQA). The student model (MobileNet-IQA) shares the same framework, but takes MobileNet as backbone.
  • Figure 3: Knowledge distillation process. MSE loss is used to minimize the discrepancy between the Student Opinion Features and the Teacher Opinion Features.
  • Figure 4: (A), (B) and (C) represent the cosine similarities of pairwise MALs within the MobileViT-IQA, MobileNet-IQA, and between MobileViT-IQA and MobileNet-IQA.
  • Figure 5: Attention maps produced by different MALs. The number of MALs is set to 3.
  • ...and 1 more figures