X-ray Image Reconstruction From Limited Angle Data For Integrated Circuit Packaging

With the rapid development of Integrated Circuit (IC) technology in the direction of highdensity and miniaturization, IC defects, especially the internal defects of packagedcomponents put forward higher requirements for detection. The emergence of ComputedTomography (CT) technology makes it possible to detect the internal defects in IC packagedcomponents. Without destroying the structure of the integrated circuit, CT can reconstruct theinner image of integrated circuit so that we can detect the inner defects effectively.However, in the actual process of IC defect detection, projection data acquired is usuallylimited by some conditions, and we only get projection data in limited angle. For instance,projection view is sparse (few-view projection) or limited within a certain range(limited-angle projection). At that time, many traditional image reconstruction methods suchas filter back projection algorithm can’t get high quality reconstruction image. With theinternal defect detection of integrated circuit packaged components as background, this paperconducts a study on the image reconstruction from incomplete projection data and finishesfollowing content:1. For the image reconstruction of integrated circuit with few-view projection data, wepropose improved POCS-TV algorithm and an image reconstruction algorithm based onvariational inequality. The improved POCS-TV algorithm improves the balance factor ofPOCS-TV algorithm and speeds up the convergence rate. The image reconstruction algorithmbased on variational inequality translates image reconstruction into a variational inequalityproblem and combines projection algorithm and image smooth operation to solve thisvariational inequality. Experiment results indicate the convergence rate of improvedPOCS-TV algorithm is faster than POCS-TV algorithm. Compared to improved POCS-TValgorithm, the reconstruction algorithm based on variational inequality has faster convergencerate, less time of a single iteration and better reconstruction quality.2. For the image reconstruction of integrated circuit with limited-angle projection data,considering that recovery capability for image information of horizontal direction gradientand vertical direction gradient is not the same, we put forward initial reconstruction model and its improved model of limited-angle projection data. Initial reconstruction model onlyuses image vertical direction gradient, and we use POCS-Vertical-TV algorithm to solve themodel which is similar to POCS-TV algorithm. The improved model reconsiders the effect ofthe horizontal direction gradient and deals with horizontal direction gradient and verticalgradient direction separately. Let horizontal direction gradient assist vertical direction gradientto reconstruct. For improved model, we propose a hybrid algorithm which combinesPOCS-Vertical-TV algorithm and soft threshold filtering algorithm. Experiment results showthe improved model and hybrid algorithm are suitable for the fan beam projection and can getgood reconstruction quality with limited-angle projection data.