Super-Sparse Projection Reconstruction Of Computed Tomography Image Based On L0Method

X-Ray examination causes more damage on human health with theincreasing of the amount of X-Ray irradiation, especially for fetus andbabies. It is not sure about that how much influence X-Ray has onhereditary, it is better for less X-Ray examination from the point ofprevention. In the process of computed tomography imaging, it is a veryuseful method to reduce X-Ray dosage using sparse projections to target.However, sparse projections will reduce necessary data and decreaseimaging quality, image reconstruction with sparse projections is alsodifficult task.There are two main methods in CT image reconstruction field—analytical method and iteration method. Filtered Back-Projection(FBP) isone of typical analytical methods, the algorithm is fast with high time anddensity resolution, however, it needs strict data, and it will result in severeartifact. While Algebraic Reconstruction Technique(ART) is one of typicaliteration algorithm, it can reconstruct image even with missing data,however, the algorithm will spend a lot of time and space, and it isdifficult to achieve it.Compressed sensing (CS) is popular theory recently which indicatesthat if signal is compressible or is sparse in some transform, thentransformed high-dimensional signal can be projected into lowerdimension through a measurement matrix which is not relevant withtransform basis, and then the original signal can be reconstructed with highprobability with solving an optimization problem. It can be proved that theprojections contain enough information. Therefore, it is meaningful to introduce CS into CT image reconstruction because CS theory providesaccurate theory support to sparse projection reconstruction.Essentially, the problem of super sparse projection reconstruction isincomplete data reconstruction. The number of unknown quantity is farmore than independent equations. Based on compressed sensing, this paperproposes a new method named approximating L0pseudo norm that is usedin CT image reconstruction. Detail process of approximating L0norm andcorresponding iteration are also proposed. At the same time, this paperusing faster methods to calculate projection matrix during algebraiciteration technique, this paper also improves Total Variation (TV) byadding weights to it. As comparisons to L0method, ART with TV methodis also designed.The paper conducts computer simulations with MATLAB, and thenanalysis the results to achieve conclusions. Five methods named L0,reweighted TV, ART-TV, ART and FBP are shown in the paper. Bothsimulation and real data experiments and STD curve show that theproposed L0method outperforms the four other methods with lessprojections, and performs comparably with other methods, so theapplication of the proposed reconstruction algorithm may permit reductionof the radiation exposure without trade-off in imaging performance.