CrossScore: Towards Multi-View Image Evaluation and Scoring

TL;DR

CrossScore is a new neural network-based image quality assessment method that compares images against multiple views without needing ground truth references, improving evaluation in tasks like novel view synthesis.

Contribution

It introduces a cross-reference assessment approach using multi-view comparisons and a novel data pipeline, filling a gap in existing image quality metrics.

Findings

Closely correlates with SSIM in quality assessment

Does not require ground truth references

Effective in novel view synthesis scenarios

Abstract

We introduce a novel cross-reference image quality assessment method that effectively fills the gap in the image assessment landscape, complementing the array of established evaluation schemes -- ranging from full-reference metrics like SSIM, no-reference metrics such as NIQE, to general-reference metrics including FID, and Multi-modal-reference metrics, e.g., CLIPScore. Utilising a neural network with the cross-attention mechanism and a unique data collection pipeline from NVS optimisation, our method enables accurate image quality assessment without requiring ground truth references. By comparing a query image against multiple views of the same scene, our method addresses the limitations of existing metrics in novel view synthesis (NVS) and similar tasks where direct reference images are unavailable. Experimental results show that our method is closely correlated to the full-reference metric SSIM, while not requiring ground truth references.