Burst Image Quality Assessment: A New Benchmark and Unified Framework for Multiple Downstream Tasks

TL;DR

This work introduces Burst Image Quality Assessment (BuIQA), a task-driven framework to score the quality of individual frames within burst sequences for downstream tasks. It builds BI-OQA and BI-SQA datasets to capture objective and subjective quality across diverse tasks, revealing that frame importance varies by task and that frame quality annotations are robust to sequence length. The proposed solution combines a Task-Driven Prompt Generation module with heterogeneous knowledge distillation and a task-aware Quality Assessment Network that uses multi-scale attention guided by prompts, achieving superior results over baselines on both objective and subjective tasks and delivering practical PSNR gains when selecting frames for denoising and super-resolution. The approach provides a scalable, adaptable tool for efficient burst processing, enabling better frame selection and improved downstream image restoration with broad applicability.

Abstract

In recent years, the development of burst imaging technology has improved the capture and processing capabilities of visual data, enabling a wide range of applications. However, the redundancy in burst images leads to the increased storage and transmission demands, as well as reduced efficiency of downstream tasks. To address this, we propose a new task of Burst Image Quality Assessment (BuIQA), to evaluate the task-driven quality of each frame within a burst sequence, providing reasonable cues for burst image selection. Specifically, we establish the first benchmark dataset for BuIQA, consisting of $7,346$ burst sequences with $45,827$ images and $191,572$ annotated quality scores for multiple downstream scenarios. Inspired by the data analysis, a unified BuIQA framework is proposed to achieve an efficient adaption for BuIQA under diverse downstream scenarios. Specifically, a task-driven prompt generation network is developed with heterogeneous knowledge distillation, to learn the priors of the downstream task. Then, the task-aware quality assessment network is introduced to assess the burst image quality based on the task prompt. Extensive experiments across 10 downstream scenarios demonstrate the impressive BuIQA performance of the proposed approach, outperforming the state-of-the-art. Furthermore, it can achieve $0.33$ dB PSNR improvement in the downstream tasks of denoising and super-resolution, by applying our approach to select the high-quality burst frames.

Figures (6)

• Figure 1: The illustration of the proposed BuIQA task and our unified framework based on task-aware prompt tuning.
• Figure 2: (a) A burst sequence from our BI-OQA dataset. Blue values indicate PSNR variations when the corresponding frame is excluded from the input sequence. (b) The boxplots and density curves of the PSNR variations when removing a frame from the sequence. The boxes and curves in different colors present the results of different sequences, and the black dots are outliers.
• Figure 3: (a) Swimlane diagrams of sequence length and quality ranking, in which different colors represent the specific burst frames. (b) Histogram of PLCC among quality scores generated by different models and downstream tasks.
• Figure 4: Illustration of our proposed framework.
• Figure 5: The illustration of our knowledge distillation.