Optical-digital joint optimization transforms the imaging paradigm from a purely optical system to an integrated framework combining an optical system with image processing algorithms, thereby significantly simplifying system architecture or enhancing system performance. In recent years, this approach has attracted increasing attention and found widespread applications. During optical-digital joint optimization, image datasets are typically selected, and the similarity between the output images and the input images is used as the loss function to guide parameter updates. However, such metrics are inherently dataset-dependent and therefore cannot objectively characterize the performance of an optical-digital imaging system. To address this limitation, this paper proposes an image quality evaluation metric based on a normalized spatial-frequency contrast sensitivity function (CSF)-weighted modulation transfer function (MTF). This metric is independent of datasets and can approximately linearly reflect the core objective of optical-digital joint optimization, namely, the similarity between the output and input images. Building upon this metric, a tolerance analysis method for optical-digital imaging systems is further proposed. Through Monte Carlo simulations, the influence of individual tolerance terms on system performance is first analyzed, followed by a comprehensive evaluation of the combined effects of all tolerance terms, enabling accurate allocation of distinct error budgets to different tolerances. In addition, an image-plane compensation algorithm applicable to non-ideal imaging systems is introduced, which effectively corrects certain aberrations induced by tolerances by adjusting the image-plane position. Collectively, these contributions provide strong support for the practical engineering implementation of optical-digital joint optimization.
Chen et al. (Fri,) studied this question.