A new approach combining a discrete iterative reconstruction-reprojection algorithm (DIRR) with a finite impulse response (FIR) low-pass filter is proposed for reconstructing images comprising opaque objects in optical computed tomography (CT). This filter, employing a rectangular window function and various bandwidths, is adopted to smooth the reprojection data between iterative reconstruction and reprojection stages. Compared reconstruction results of the traditional iterative reconstruction-reprojection (IRR) algorithm, projection space iteration reconstruction-reprojection (PSIRR) algorithm, and the new DIRR algorithm for an asymmetrical four-peak and a single-peak test image, including a circle round opaque object, are studied. The results show that the reconstruction precision of the new algorithm is related to the value of the bandwidth of the finite impulse response (FIR) filter, and the optimal bandwidth increases when the space frequency of the reconstructed image ascends. Furthermore, the DIRR algorithm with an optimal bandwidth has an obviously higher reconstruction precision than IRR and PSIRR algorithms, and has a potential application of reconstructing images, including obstacle objects with limited views.