Since the inception of computed tomography (CT) imaging, researchers had started striving to meet its intrinsic objectives. One of the central objectives is to achieve low dose/incomplete sinogram reconstructions with high image quality. An incomplete sinogram is procured either by reducing projection views or by lowering the x-ray tube current. In the former case, the reconstructions will consist of aliasing artefacts. In the latter case, photon starvation occurs and resultsin electronic noise. Electronic noise introduces noisy measurements (zero and nonpositive values) in the sinogram. Any corrections to these values cause deformed reconstructions. This work introduces new and improved versions of conventional filter back projection (FBP) and iterative reconstructions (IR) algorithms. These algorithms employ computer vision and machine learning concepts to deal with aliasing artefacts and electronic noise. All the new algorithms reconstructed 3D images without correcting the noisy measurements and achieved 96-100% Structural SIMilarity (SSIM) index. Furthermore, when tested on spongy and porous materials, resultant reconstructions achieved 96% and 100% accuracy in SSIM and porosity analyses, respectively. Besides, these new algorithms are computationally efficient since they reconstruct 3D images with incomplete sinograms reducing the data acquisition and reconstruction time by 42-75%.