Dual-energy CT Reconstruction using Guided Image Filtering

Dual-energy computed tomography (CT) has become widely applied due to its advantages like reducing beam-hardening artefacts and providing more accurate attenuation measurements. Compared with traditional CT (using single X-ray spectrum), dual-energy CT performs much better especially when some material cannot be easily distinguished by one X-ray spectrum. When doing dual-energy CT reconstruction, it is a basic principal to fully exploit the information of the data from both high and low energies. Here we propose a novel dual-energy CT reconstruction method using the guided image filtering algorithm and the commonly used transmission image reconstruction method: separable quadratic surrogates (SQS). As a patch-based image processing algorithm, guided image filtering has nice properties of edge-preserving smoothing and relatively high computational efficiency. It can be used in structure-transferring between two images, filtering-based feathering/matting and dehazing. On the other hand, the SQS algorithm is an ordered subsets algorithm for penalty likelihood image reconstruction in transmission tomography. It can rapidly decrease the objective function value in the early iterations. In this paper, the ideas of guided image filtering and SQS are explained. We study the basic principle and main advantages of guided image filtering and SQS, and tell how we combine these two methods to do dual-energy CT reconstruction. After that, we did computed simulation (with simulated Poisson noise) to validate our idea, and make comparison with other methods. The results show that the quality of both images can be improved in a satisfying convergence speed.