Abstract:Image quality assessment (IQA) is very important for both end-users and service providers since a high-quality image can significantly improve the user's quality of experience (QoE) and also benefit lots of computer vision algorithms. Most existing blind image quality assessment (BIQA) models were developed for synthetically distorted images, however, they perform poorly on in-the-wild images, which are widely existed in various practical applications. In this paper, we propose a novel BIQA model for in-the-wild images by addressing two critical problems in this field: how to learn better quality-aware feature representation, and how to solve the problem of insufficient training samples in terms of their content and distortion diversity. Considering that perceptual visual quality is affected by both low-level visual features (e.g. distortions) and high-level semantic information (e.g. content), we first propose a staircase structure to hierarchically integrate the features from intermediate layers into the final feature representation, which enables the model to make full use of visual information from low-level to high-level. Then an iterative mixed database training (IMDT) strategy is proposed to train the BIQA model on multiple databases simultaneously, so the model can benefit from the increase in both training samples and image content and distortion diversity and can learn a more general feature representation. Experimental results show that the proposed model outperforms other state-of-the-art BIQA models on six in-the-wild IQA databases by a large margin. Moreover, the proposed model shows an excellent performance in the cross-database evaluation experiments, which further demonstrates that the learned feature representation is robust to images with diverse distortions and content. The code is available at https://github.com/sunwei925/StairIQA.

Multiple Level Feature-Based Universal Blind Image Quality Assessment Model.

Multitask Convolutional Neural Network for No-Reference Image Quality Assessment

SCVS: Blind Image Quality Assessment Based on Spatial Correlation and Visual Saliency

Deep Blind Image Quality Assessment Using Dynamic Neural Model with Dual-order Statistics

McmIQA: Multi-Module Collaborative Model for No-Reference Image Quality Assessment

Blind Image Quality Assessment Using a Deep Bilinear Convolutional Neural Network

Deep Supervised Dictionary Learning for No-Reference Image Quality Assessment

A Multiscale Approach to Deep Blind Image Quality Assessment

Blind 3D Image Quality Assessment Based on Multi-scale Feature Learning

No-reference image quality assessment with multi-scale weighted residuals and channel attention mechanism

Blind Image Quality Assessment Via Cross-View Consistency

LIQA: Lifelong Blind Image Quality Assessment

Deep blind image quality assessment based on multiple instance regression

Multi-task Feature Enhancement Network for No-Reference Image Quality Assessment

Adaptive Mixed-Scale Feature Fusion Network for Blind AI-Generated Image Quality Assessment

AM-BQA: Enhancing blind image quality assessment using attention retractable features and multi-dimensional learning

Blind Quality Assessment for in-the-Wild Images via Hierarchical Feature Fusion and Iterative Mixed Database Training

Hierarchical Curriculum Learning for No-Reference Image Quality Assessment

An Accurate Deep Convolutional Neural Networks Model for No-Reference Image Quality Assessment

Opinion-Unaware Blind Image Quality Assessment using Multi-Scale Deep Feature Statistics

Deep Superpixel-Based Network For Blind Image Quality Assessment