LIQA: Lifelong Blind Image Quality Assessment

Abstract

Existing learning-based methods for blind image quality assessment (BIQA) are heavily dependent on large amounts of annotated training data, and usually suffer from a severe performance degradation when encountering the domain/distribution shift problem. Thanks to the development of unsupervised domain adaptation (UDA), some works attempt to transfer the knowledge from a label-sufficient source domain to a label-free target domain under domain shift with UDA. However, it requires the coexistence of source and target data, which might be impractical for source data due to the privacy or storage issues. In this paper, we take the first step towards the source-free unsupervised domain adaptation (SFUDA) in a simple yet efficient manner for BIQA to tackle the domain shift without access to the source data. Specifically, we cast the quality assessment task as a rating distribution prediction problem. Based on the intrinsic properties of BIQA, we present a group of well-designed self-supervised objectives to guide the adaptation of the BN affine parameters towards the target domain. Among them, minimizing the prediction entropy and maximizing the batch prediction diversity aim to encourage more confident results while avoiding the trivial solution. Besides, based on the observation that the IQA rating distribution of single image follows the Gaussian distribution, we apply Gaussian regularization to the predicted rating distribution to make it more consistent with the nature of human scoring. Extensive experimental results under cross-domain scenarios demonstrated the effectiveness of our proposed method to mitigate the domain shift.

Semi-Supervised Blind Image Quality Assessment through Knowledge Distillation and Incremental Learning

“…Incremental or continual learning research in the field of BIQA is still in its nascent stage. LwF-KG (Liu et al 2022) pioneered the concept of continual learning in BIQA and introduced a simple yet effective approach. By building upon a shared backbone network, they appended a prediction head for a new dataset and imposed a regularizer, enabling all prediction heads to evolve with new data while mitigating catastrophic forgetting of old data.…”

“…With the development of deep learning, many researchers finetuned the pre-trained models to solve the insufficient data problem. CNN-based BIQA methods (Kang et al 2014;Liu et al 2022) directly used or fine-tuned a pretrained CNN classification model as a feature extractor to further predict image quality scores. Ke et al (Ke et al 2021) used Vision Transformer as the backbone for feature extraction.…”

Subjective Assessment of Objective Image Quality Metrics Range Guaranteeing Visually Lossless Compression

“…Ying et al [ 36 ] prepared a dataset of distorted images, patches, and subjective qualities, and applied a deep learning-based model in order to predict image quality scores. An adaptive blind IQA framework proposed by Liu et al [ 37 ] utilizes a variety of distortion and quality grades to generate pseudo features. Similarly, Zhang et al [ 38 ] presented an accurate and stable continual learning-based approach trained on different IQA databases.…”

Using HVS Dual-Pathway and Contrast Sensitivity to Blindly Assess Image Quality

“…Experimental results show the effectiveness of the proposed method compared to standard training techniques for BIQA. Liu et al [ 42 ] proposed a lifelong IQA (LIQA) method to address the challenge of adapting to unseen distortion types by mitigating catastrophic forgetting and learning new knowledge without accessing previous training data. It utilizes the Split-and-Merge distillation strategy to train a single-head network for task-agnostic predictions.…”