目的 图像的内插处理中,常将图像分解成许多子块,在块与块的交界处或图像边缘会引起图像质量的损伤。为了克服边缘效应,本文提出基于2维全相位数字信号处理的内插算法。方法 首先设计加窗的2维全相位数字信号处理算法,并利用该算法得到基于离散傅里叶变换(DFT)、离散余弦变换(DCT)和离散小波变换(DWT)的3种类型2维图像内插核,可同时处理行列及斜向频率分布。其次,为保持线性相位特性,对2维基窗和特性矩阵H的构成提出约束条件。结果 该方法在有效地降低分块效应的同时,解决了对斜向内容分量无法有效操作的问题,使内插得到的图像PSNR至少提升3 dB。结论 设计针对不同频率分布的3种图像内插实验,并同传统6种内插方式结果进行对比,表明该算法物理意义明确,恢复效果良好,实用性较强。

Objective Traditional image interpolation methods separate an image into many sub-blocks. In this case, the quality of the recovered image becomes poor, particularly in the boundary of the blocks and the image borders. To eliminate this side effect, we present a new algorithm based on the 2D all-phase digital signal processing method.Method First, an algorithm for 2D all-phase windowed processing is designed to achieve three kinds of interpolation kernel based on DFT, DCT, and DWT. The three kernels, which are built directly from the 2D transmission property H, can process the frequency distribution along the row and column as well as along the diagonal slope directions. Both H and the base window should be confined for phase linearity.Result The proposed algorithm effectively eliminates the blocking effects and solves the uncontrollable contents in the diagonal slope directions using a traditional technique. In addition, the proposed algorithm improves the PSNR of the recovered images by at least 3 dB.Conclusion Comparison experiments on the proposed algorithm and six traditional methods show that the proposed algorithm is physically explicit and applicable.