Skip to Main content Skip to Navigation

Scattering correction in cone beam tomography using continuously thickness-adapted kernels

Abstract : Advanced Cone Beam Computed Tomography (CBCT) typically uses a divergent conebeam source and a large area detector. As a result, there an inevitable increase in the size area of illumination causing an increase in the intensity of X-ray scatter signal, both from the object and the detector. This leads to the violation of prime assumption of reconstruction process which is based on straight line integrals path followed by the photons. Consequently scatter artifacts appear in the reconstruction images as steaks, cupping effect and thus produce wrong reconstruction values. Due to the severity of the reconstruction artifact caused by scatter, many scatter corrections methods have been adopted in literature. The first part of this study, reviews most of the existing scatter correction methods. The effect of scattering becomes more prominent and challenging in case of X-ray source of high energy which is used in industrial Non Destructive Testing (NDT), due to higher scatter to primary ratio (SPR). Therefore, in this study, we propose a continuously thickness-adapted deconvolution approach based on improvements in the Scatter Kernel Superposition (SKS) method. In this method, the scatter kernels are analytically parameterized over the whole thickness range of the object under study to better sample the amplitude and shape of kernels with respect to the thickness. The method is tested for both homogeneous and heterogeneous objects as well as simulated and experimental data. Another important aspect of this study is the comprehensive evaluation of contribution of the detector scatter performed using continuous method by separating the contribution of scatter due to the object and the detector. This is performed by modeling the scatter kernels using a four-Gaussian model. In the first approach, we performed this evaluation based on simulation of kernels from Monte Carlo simulations and the corrections are performed on typical industrial experimental data. The results obtained prove that the scatter correction only due to the object is not sufficient to obtain reconstruction image, free from artifacts, as the detector also scatters considerably. In order to prove this point experimentally and to have a better modeling of the detector, we describe a method based on combination of experiments and simulations to calculate the scatter kernels. The results obtained also prove, the contribution of the detector scattering becomes important and the PSF of the detector is not constant as considered in the studies so far, but it varies to a great extend with the energy spectrum.
Document type :
Complete list of metadatas

Cited literature [95 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Friday, January 26, 2018 - 4:37:09 PM
Last modification on : Saturday, October 17, 2020 - 3:20:19 AM
Long-term archiving on: : Friday, May 25, 2018 - 10:58:44 AM


Version validated by the jury (STAR)


  • HAL Id : tel-01693958, version 1


Navnina Bhatia. Scattering correction in cone beam tomography using continuously thickness-adapted kernels. Image Processing [eess.IV]. Université de Lyon, 2016. English. ⟨NNT : 2016LYSEI092⟩. ⟨tel-01693958⟩



Record views


Files downloads