cs.CVAug 27, 2026

Rethinking Image Processing for the Age of AI: A Problem-First Framework for Scientific Progress

Authors: Guoping Qiu

Organizations: University of Nottingham Ningbo China, Ningbo, China · University of Nottingham, Nottingham, United Kingdom

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Modern AI has greatly expanded the capabilities of image processing. However, the ready availability of powerful models, public datasets, and benchmark leaderboards has also en- couraged a model-first research pattern: researchers increasingly begin with an available architecture and optimize it on a public benchmark, rather than beginning with the underlying real-world imaging problem. This can produce impressive benchmark results without necessarily improving our understanding or solution of the real problem. This paper argues for a problem-first approach that distinguishes the physical imaging problem, solution principle, statistical estimator, and computational implementation, while clarifying what modern AI can achieve and which fundamental problems remain unsolved. Through case studies of super- resolution and low-light enhancement, we show how benchmark datasets may define tasks that differ substantially from the real-world problems they are intended to represent, and why performance improvements must be interpreted within the conditions under which they are obtained. We propose a six-stage workflow that places problem formulation, image acquisition, information-loss analysis, assumptions, ambiguity, and evaluation before model and dataset selection. The paper also proposes clearer standards for evidence, reproducibility, uncertainty, and claims of state-of-the-art performance. More fundamentally, it calls for a change in research culture and education so that future researchers learn to understand imaging problems deeply and use modern AI to achieve genuine scientific and technical advancement.

Figures & tables

LevelCentral questionTypical choicesTypical failure when conflated
ProblemWhat latent quantity should be inferred from what observation, under which acquisition and use?Sensor-to-radiance recovery, canonical rendering, perceptual enhancement, recognition-oriented preprocessingA benchmark construction silently becomes the real-world problem
PrincipleWhat knowledge makes the problem solvable?Bayesian inference, regularization, empirical risk minimization, invariance, self-consistencyA generic learning principle is presented as a problem-specific explanation
EstimatorWhich decision is optimal for the loss and uncertainty?Conditional mean, median, MAP estimate, posterior sample, robust decisionDifferent outputs are ranked as if they estimated the same target
ImplementationHow is the estimator represented and computed?Filter, optimization, CNN, transformer, diffusion sampler, hybrid systemArchitectural novelty is mistaken for new information or new physics
TABLE I: Four levels that should be reported separately in an image-processing study.
DimensionDesigned/model-based processingLearning by exampleDeep and generative learning
RepresentationUsually designed; assumptions visible in filters, transforms, and regularizersDesigned features with learned predictors, dictionaries, exemplars, or priorsRepresentation and processing learned jointly at high capacity
Primary knowledgePhysics and analytic assumptions ( Kp,KaK_{p},K_{a} )Internal or external examples ( Ki,KeK_{i},K_{e} ), often with explicit structureLarge external priors, architecture bias, and increasingly physics/internal adaptation
Typical inferenceDirect filtering or explicit optimizationRegression, lookup, retrieval, or sparse inferenceEnd-to-end regression, unfolding, adversarial inference, or posterior sampling
Main strengthTransparency, data efficiency, predictable constraintsEmpirical flexibility with moderate interpretabilityCapacity, context, transfer, and complex or multimodal conditionals
Characteristic riskModel bias and restrictive priorsLimited examples or representation capacityDistribution shift, opaque priors, and convincing unsupported detail
What remains invariantObservation formation, information loss, identifiability, loss-dependent decisions, and the need to justify prior information
TABLE II: A problem-first comparison of three broad image-processing eras. The boundaries are porous; these are dominant tendencies rather than exclusive properties.
Evidence componentWhat it establishesWhat it does not establish alone
Matched synthetic benchmarkControlled accuracy for a known degradation and referenceRobustness to unknown acquisition or real sensor pipelines
Captured paired benchmarkPerformance for a specific capture and registration protocolUniversal real-world generalization or unique ground truth
Cross-dataset/sensor transferSensitivity to dataset and device shiftCoverage of all deployments
Forward-model stress testBehaviour under controlled physical parameter changesAdequacy of the chosen model family
Re-degradation consistencyCompatibility of output with observation under an assumed modelTruth of prior-dominated null-space detail
Unpaired in-the-wild testingQualitative failure modes and deployment diversityFull-reference fidelity
TABLE III: A portfolio of evidence for claims about an imaging method.
Community actorConstructive practiceCultural signal
Students and researchersBegin with target, acquisition, information loss, assumptions, and use; treat architecture as a justified design decisionUnderstanding the problem is part of technical excellence
Supervisors and research leadersRequire a problem statement and falsifiable claim before model development; discuss failures and mismatch, not only average scoresCareful formulation and diagnosis are legitimate research outputs
AuthorsMatch claims to evidence; disclose resources and tuning; report robustness, uncertainty, and limitationsCredibility matters more than rhetorical breadth
ReviewersEvaluate whether the method solves the stated problem and whether comparisons are fair; recognize non-architectural noveltyThe highest familiar-benchmark score is not the only form of contribution
Editors and programme committeesBroaden scope and reviewer guidance; welcome problem, data, measurement, evaluation, and diagnostic papersPublication rewards cumulative knowledge, not only leaderboard movement
Benchmark organizersDocument construction; refresh tests; include shifts, efficiency, uncertainty, and worst-group resultsBenchmarks are instruments for inquiry rather than definitions of reality
TABLE IV: Contributions to a problem-first image-processing culture.
More in this topic

Benchmark Design

Benchmark design creates tasks, datasets, metrics, and protocols that make AI capabilities measurable and comparisons credible.

See all 394 papers

Explore similar work

CardsList
Jul 8, 2026

Does AI Understand Imaging? A Systematic Benchmark of Agentic AI for Computational Imaging Tasks

Vision-language models (VLMs) and agentic AI have shown strong performance on semantic visual tasks, but it remains unclear whether they can handle the physics and inverse problems that underlie computational imaging. We present ImagingBench, a benchmark of 20 computational imaging tasks spanning five categories: ray and wave optics, image signal processing, inverse reconstruction, computational sensing, and calibration. ImagingBench evaluates three complementary settings: Expert, fixed expert-guided inverse reconstruction; Planner, planner-guided inverse reconstruction; and Forward, forward-system simulation for consistency checking. We benchmark leading proprietary and open-source image-centric multimodal systems, including Gemini, GPT, and Qwen, and compare them with representative task-specific non-agentic baselines. Across tasks, agentic models remain consistently weaker than specialized methods, especially on computational sensing problems such as lensless imaging, event-based reconstruction, time-of-flight imaging, and holography. Planner guidance provides only modest and inconsistent gains over the fixed-prompt Expert baseline. Although the models often generate visually plausible outputs, their reference-based fidelity remains poor, revealing a substantial gap between semantic visual competence and physically grounded imaging performance. ImagingBench provides a unified testbed for measuring this gap and tracking progress in agentic AI for computational imaging.

Ethan Chung, Chuanjun Zheng, Jasper Tan +3

University of Hawaii at Manoa · Glass Imaging

VLM EvaluationAI Agent BenchmarksComputational ImagingInverse Problems
AskPDF
Jul 12, 2026

Imaging-101: Benchmarking LLM Coding Agents on Scientific Computational Imaging

Computational imaging, which recovers hidden signals from indirect, noisy measurements, underpins quantitative discovery across scientific disciplines, yet building a correct reconstruction pipeline demands deep domain expertise and remains laborious even for domain scientists. We introduce Imaging-101, a benchmark of 57 expert-verified computational imaging tasks spanning six scientific domains, each grounded in a peer-reviewed paper and canonicalized into a standardized four-stage pipeline (preprocessing, forward physics modeling, inverse solver, and visualization) Three evaluation tracks (planning, function-level unit tests, and end-to-end reconstruction) probe distinct agent capabilities across the full pipeline. Evaluating seven frontier LLMs uncovers systematic challenges in applying coding agents to computational imaging that go beyond those exposed by general coding benchmarks, spanning algorithm selection, physical convention handling, and pipeline integration. These findings highlight concrete capability gaps and point toward skill-augmented, domain-specialized agents as a practical path to reliable computational imaging assistance.

Siyi Chen, Jiahe Ying, Yixuan Jia +9

Key Laboratory of Spacecraft Design Optimization and Dynamic Simulation Technologies, Ministry of Education, Beijing, 102206, China

AI Agents for Scientific DiscoveryComputational ImagingLLM AgentsImage Inverse Problems
AskPDF
Oct 8, 2026

OpenProblemBench: Benchmarking AI on Open Problems in the Foundational Theoretical Sciences

The next frontier for artificial general intelligence is tackling unresolved scientific problems, calling for benchmarks that assess progress beyond established knowledge. We introduce OpenProblemBench, a benchmark of 82 unresolved problems drawn from the mathematics and theoretical physics literature. Each problem supplies the research context, assumptions, and prior progress needed to investigate the question. We select problems whose proposed solutions admit comparatively clear checks of their decisive mathematical or computational claims. Four evaluator models independently assess the correctness, completeness, and degree of progress of each submission without reference solutions. Across seven evaluated configurations, GPT-6-Astra achieves the highest mean judged solve rate of 14.0%, compared with 5.5-6.7% for the evaluated full-size open models and 2.4-3.7% for Flash models. Case comparisons connect stronger outcomes to changes in problem representation, general arguments that extend beyond finite evidence, and proofs of the steps needed to complete a solution. By grounding evaluation in questions arising from the research literature, OpenProblemBench provides a setting for investigating the capabilities and limitations of AI as a contributor to foundational theoretical science.

Zhiyi Li, Sihan Hu, Tianning Xiao +4

Department of Modern Physics, University of Science and Technology of China, Hefei, Anhui 230026, China · CAS Key Laboratory of Theoretical Physics, Institute of Theoretical Physics, Chinese Academy of Sciences, Beijing 100190, China · Hefei National Laboratory, University of Science and Technology of China, Hefei 230088, China +3

AI for ScienceScientific Reasoning in Language ModelsMathematical Reasoning Benchmarks
AskPDF