在相控阵超声波无损检测领域，全矩阵捕获(FMC)以及相应的全聚焦后处理算法(TFM)具有优异的成像质量。但是全聚焦算法需要耗费大量的计算资源，尤其对于分层物体的检测。本文对全聚焦算法成像过程进行了深入研究，针对二维单一介质物体和分层物体的检测对全聚焦算法进行改进，极大提升算法的性能，并将优化策略扩展至三维物体的超声波全聚焦成像算法。本文的主要贡献包括以下工作：1.本文将光栅图形学中的直线扫描转换算法引入全聚焦过程，提出了基于直线扫描转换的全聚焦成像算法(LSC-TFM)。LSC-TFM利用直线扫描转换算法获得位于直线上的像素点坐标，使用加法运算替换原有的均方根运算，极大减少了均方根运算的数量。基于相同的优化策略，将该算法扩展至三维物体的超声检测，实验证明，LSC-TFM的成像速度为原算法的2倍。2.基于光栅图形学中的椭圆扫描转换算法，提出了基于椭圆弧扫描转换的二维全聚焦算法(ESC-TFM)，能够将成像过程转换为多次椭圆弧扫描转换操作。ESC-TFM通过去除大量耗时的均方根运算，并有效利用全矩阵数据的稀疏性来提升成像速度。IESC-TFM在ESC-TFM的基础上进行改进，通过坐标平移获得椭圆弧上像素点坐标，从而减少了椭圆扫描转换算法的调用次数。经实验验证，IESC-TFM成像速度相对于原算法提升了14倍。3.将ESC-TFM扩展至三维物体的超声成像过程，提出了基于椭球面扫描转换的三维全聚焦算法(SCES-TFM)。在SCES-TFM中，整个成像过程转化为多次椭球面扫描转换操作。ISCES-TFM在SCES-TFM的基础上进行修改，对于相同形状的椭球面，可使用坐标平移获得椭球面上像素点坐标，使椭球面扫描转换算法的调用次数降低了一个数量级。实验表明，ISCES-TFM成像速度为原算法的8倍。4.在分层物体的全聚焦成像过程中，位于同一传输路径上的像素点，其所对应的折射点相同。因此本文将折射点的位置看作已知条件，而将目标像素点位置看作未知条件，从而避免了原光线跟踪算法中大量的迭代运算。通过将结果图像划分为多个成像区域，提出了基于区域划分的二维全聚焦算法(2D RD-TFM)。实验证明，RD-TFM成像速度相对于原算法有一个数量级的提升。5.将区域划分算法扩展至三维分层物体的超声成像过程中，提出了基于区域划分的三维全聚焦算法(3D RD-TFM)。3D RD-TFM通过将不同介质分界面上的相邻像素点划分为多个三元组实现了三维结果图像的区域划分。经实验验证，3D RD-TFM运行时间为原算法的1/50。

In phased array ultrasonic testing, Full Matrix Capture (FMC) and its post processing algorithm total focusing method (TFM) is favored by numerous researchers due to their superior image quality. However the extremely large amounts of ultrasonic data require lots of computing resources, especially for multi-layered objects. This dissertation provides an in-depth understanding of the TFM algorithm in ultrasonic testing of both homogeneous and heterogeneous 2D objects, which greatly improves its performance. The optimization strategy is also extended to 3D object. The main contributions of this dissertation are listed as follows:(1) By introducing the linear scan conversion algorithm in raster graphics, the total focusing method based on linear scan-conversion (LSC-TFM) is proposed. The linear scan-conversion algorithm are used to calculate the coordinates of pixels in the same line. The distances between pixels and elements are calculated using addition instead of RMS. Therefore, lots of RMS computation are avoids. This optimization strategy still applies to 3D ultrasonic imaging. The simulation experiment proves that the imaging speed of LSC-TFM is twice as fast as the traditional TFM.(2) Based on the elliptical scan conversion algorithm in raster graphics, the total focusing method based on elliptical scan conversion (ESC-TFM) are proposed, which converts the imaging process to multiple scan-conversion operations of elliptical arcs. The performance of TFM is improved because it avoids time-consuming RMS computations and fully utilizes the sparsity of the acquired full matrix. Based on ESC-TFM, IESC-TFM uses coordinate translation to acquire coordinates of pixels on an elliptical arc, which greatly reduces the calling number of the elliptical scan conversion algorithm. The experiment verifies that the imaging speed of ESC-TFM is improved 14 times.(3)Based on the same optimization strategy, ESC-TFM is extended to 3D ultrasonic imaging, which is 3D total focusing method based on Scan Converting Ellipsoidal Surface (SCES-TFM). In SCES-TFM, the imaging process is divided into multiple scan-conversion operations of ellipsoidal surfaces. Besides, based on SCES-TFM, ISCES-TFM uses coordinate translation to acquire coordinates of pixels on a given ellipsoidal surface, which reduces the calling number of ellipsoidal surface scan-conversion algorithm by one order of magnitude. The experiment shows that the imaging speed of ISCES-TFM is 8 times as fast as the conventional 3D TFM.(4) In ultrasonic imaging of multi-layered objects, every pixel on the same propagation path shares the same incident point. In this paper, the location of POI is considered as known condition and the target pixel as unknown condition, which avoids lots of iterative computations in ray-tracing method. By dividing the resulting image into multiple regions, the total focusing method based on region-division (2D RD-TFM) is presented. According to the experiment, the imaging speed of RD-TFM is increased by one order of magnitude.(5)By extending RD-TFM to 3D ultrasonic imaging of multilayered object, 3D total focusing method based on region division (3D RD-TFM) are proposed. The pixels at the interface between different media are divided into multiple tuples where each tuple contains three adjacent pixels. The entire 3D image is divided into multiple regions where each region corresponds to a tuple. The experiment shows that the processing time of 3D RD-TFM is only 1/50 of the conventional 3D TFM with ray-tracing.