TRIQA: Image Quality Assessment by Contrastive Pretraining on Ordered Distortion Triplets

2025-07-16

AI Summary

To address poor model generalizability in no-reference image quality assessment (NR-IQA) caused by scarcity of subjective quality annotations, this paper proposes a contrastive pretraining framework based on ordered distortion triplets. The method requires only a small set of pristine reference images to construct content-aware, customized triplet datasets. It introduces a novel contrastive learning objective that jointly enforces content invariance and quality sensitivity, integrated with a quality-aware network architecture for end-to-end optimization. Evaluated on benchmark NR-IQA datasets—including LIVE and CSIQ—the approach significantly outperforms existing few-shot NR-IQA methods. Notably, it maintains high accuracy and strong generalization even when trained on triplets derived from as few as ten source images. This work establishes a new paradigm for perception-driven quality assessment under low-resource conditions, enabling effective NR-IQA with minimal supervision.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Learning to Search
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Image Quality Assessment (IQA) models aim to predict perceptual image quality in alignment with human judgments. No-Reference (NR) IQA remains particularly challenging due to the absence of a reference image. While deep learning has significantly advanced this field, a major hurdle in developing NR-IQA models is the limited availability of subjectively labeled data. Most existing deep learning-based NR-IQA approaches rely on pre-training on large-scale datasets before fine-tuning for IQA tasks. To further advance progress in this area, we propose a novel approach that constructs a custom dataset using a limited number of reference content images and introduces a no-reference IQA model that incorporates both content and quality features for perceptual quality prediction. Specifically, we train a quality-aware model using contrastive triplet-based learning, enabling efficient training with fewer samples while achieving strong generalization performance across publicly available datasets. Our repository is available at https://github.com/rajeshsureddi/triqa.

Problem

• Predicting image quality without reference images
• Overcoming limited labeled data for NR-IQA models
• Enhancing generalization with contrastive triplet-based learning

Innovation

• Contrastive triplet-based learning for quality-aware model
• Custom dataset from limited reference content images
• Combines content and quality features for prediction