無透鏡相機中稀疏隨機二值PSF之研究

Abstract

遮罩式無透鏡相機是一種新型的成像架構，它去除傳統相機的鏡頭結構，改以光學遮罩調控光場，使被攝物的光到達感測器前經過特定的光學調變，並依靠後端的重建演算法還原影像。在相關研究中，PhlatCam是一篇具代表性的論文。該論文提出Perlin noise contour PSF作為點擴散函數 (point spread function, PSF) 的設計，在成像品質上優於過去的其他PSF設計。除此之外，論文中提到near-field phase retrival (NfPR)的演算法，使理想的PSF得以轉換成實際可製作的相位遮罩。 本研究提出稀疏隨機二值點擴散函數(sparse-random-binary PSF，SRB-PSF)，並透過NfPR演算法生成對應的相位遮罩噢光學傳遞後的實際PSF。為了完整評估旗成像能力，本研究建立無透鏡成像的電腦模擬流程，包含光學調變、雜訊引入以及正則化重建，並分別採用Tikhonov正則化與total variation (TV)正則化進行重建分析，影像品質則透過結構相似度指標(SSIM)與峰值訊噪比(PSNR)進行量化評估。 在分析過程中，本研究系統性比較不同原圖內容、不同重建流程與多組SRB-PSF設計變因對重建品質之影響，探討 PSF 稀疏程度與頻域特性對影像重建能力之關聯。模擬結果顯示，所提出之SRB-PSF在多數條件下皆能取得優於既有PSF設計的重建品質。此外，透過多組參數測試與實驗觀察，本研究也歸納出較佳 PSF 設計趨勢，可作為後續無透鏡成像PSF優化設計之參考依據。

Mask-based lensless cameras have emerged as a novel imaging architecture that eliminates conventional lenses and instead employs an optical mask to modulate the incident light field. The modulated measurements are subsequently reconstructed into images through computational algorithms. Among related works, PhlatCam is a representative study that introduced a Perlin noise contour point spread function (PSF) design, demonstrating improved imaging performance compared to previous PSF configurations. In addition, the study proposed a near-field phase retrieval (NfPR) algorithm, enabling the conversion of an ideal PSF into a physically realizable phase mask. In this work, we propose a sparse-random-binary PSF (SRB-PSF) design and utilize the NfPR algorithm to generate the corresponding phase mask and the resulting practical PSF after optical propagation. To comprehensively evaluate its imaging capability, a complete computational simulation framework for lensless imaging is established, including optical modulation, noise modeling, and regularized image reconstruction. Both Tikhonov regularization and total variation (TV) regularization are employed for reconstruction analysis. The reconstructed image quality is quantitatively evaluated using the Structural Similarity Index (SSIM) and the Peak Signal-to-Noise Ratio (PSNR). During the analysis, we systematically investigate the effects of different scene contents, reconstruction strategies, and multiple SRB-PSF design parameters on reconstruction performance. The relationship between PSF sparsity, its frequency-domain characteristics, and image reconstruction quality is examined. Simulation results demonstrate that the proposed SRB-PSF achieves superior reconstruction performance compared to existing PSF designs under most evaluated conditions. Furthermore, through extensive parameter studies and empirical observations, favorable PSF design tendencies are identified, providing useful insights for future optimization of PSF design in lensless imaging systems.