Quantitative Quality Assessment of Computational Photography (A06)

计算摄影(CP)为数字成像开辟了新的可能性，但文献中描述的方法的多样性给它们各自的表现带来了客观评估的问题。本项目建议通过以下几个步骤来解决这个问题：(1)确定一组基准场景，这些基准场景可以廉价且精确地以量化的不确定性再现，作为新方法的试金石；(2)定义、获取和分发综合光场数据集以用作真实模拟中的基准(3)开发基于基准场景的显著区域评估的CP的预测性能度量。

Computational Photography (CP) opens up novel possibilities for digital imaging, but the diversity of the approaches described in the literature pose problems regarding the objective assessment of their respective performance. This project proposes to address this by a number of steps: (1) identifying a set of benchmark scenes which can be inexpensively and precisely reproduced with quantified uncertainty to serve as litmus tests for new methods; (2) defining, acquiring and distributing comprehensive light field data sets to be used as benchmarks in realistic simulations (3) develop predictive performance metrics for CP as evaluated on salient regions of the benchmark scenes.