The need for edge detail reconstruction under low sampling rates is increasing for applications such as microscopic imaging, tomography and computer vision. Compressive sensing ghost imaging (CSGI) with undersampling greatly depends on the sparsity of the target imaged and edge reconstruction is not satisfactory. In this paper, we propose a ghost imaging (GI) reconstruction method based on total variation regularization GI (TVRGI) which takes advantage of the spatial sparsity of the target and the good edge-preserving property of total variation regularization to obtain a reconstructed image with a clearer edge. By imaging reconstructions for both binary and grayscale objects, our method presents a more satisfactory visual imaging effect performance than CSGI. It is also shown that TVRGI is more applicable for GI with grayscale objects.

对于显微成像、断层扫描和计算机视觉等应用，在低采样率下对边缘细节重建的需求正在增加。欠采样的压缩感知重影成像（CSGI）很大程度上取决于成像目标的稀疏性，边缘重建效果不佳。在本文中，我们提出了一种基于全变分正则化GI（TVRGI）的鬼成像（GI）重建方法，该方法利用目标的空间稀疏性和全变分正则化的良好边缘保留特性，获得具有更清晰的边缘。通过对二进制和灰度对象的成像重建，我们的方法呈现出比 CSGI 更令人满意的视觉成像效果性能。还表明 TVRGI 更适用于具有灰度对象的 GI。