In this research, we propose an algebraic reconstruction method suitable for the coherent detection imaging (CDI) based transillumination laser CT. When imaging highly scattering media such as tissues, the laser CT does not obey a simple absorption model, i.e., the Radon transform, because of the surface effects that arise due to refractive index mismatch at the object boundary. The surface effects degrade the reconstructed images quality. To compensate for the surface effects. However, a stable solution can not be obtained from the expression, since the equation system is always an underdetermined one. The constraint from the quantitative relationship between projections is considered to obtain the stable solution. The constraint ins based on the properties of CT and Gaussian beam. Accordingly, our reconstruction results in a least squares problem with a constraint. The problem is solved via the conjugate gradient method, whose convergence rate is relatively high. We demonstrate the effectiveness by applying the proposed method to experimental data acquired from a physical phantom.