能量归一块匹配三维协同滤波及地震去噪应用

传统的块匹配三维协同滤波方法在数字信号去噪中虽然取得较好的应用效果,但其块匹配及分组结果易受块间能量差异影响,参数选取效率不高且去噪效果有待进一步提升,为此,提出一种基于分块能量归一的块匹配三维协同滤波方法,该方法主要包括以下2步:①先对分块数据在二维变换(先在行(列)方向进行一维变换,再在列(行)方向进行一维变换)域进行软/硬阈值滤波和分块能量归一处理;②进行二维变换域块匹配及分组、块顺序方向一维变换(即分块数据的三维变换)和三维变换域软/硬阈值滤波、分块数据三维反变换和保持分块数据能量关系的块聚集。理论分析、模型数据试验和实际数据应用结果表明,该方法能消除块间能量差异的影响,提升相似块匹配精度和计算效率,最终提升去噪效果,可被广泛用于去噪处理。

The traditional block-matching 3D collaborative filtering method has achieved good results in digital signal denoising.However,its block-matching and grouping results are easily affected by the energy difference between blocks,in which case the efficiency of parameter selection is not high.Therefore,further improvement of the denoising effect is warranted.In this work,a block-matching 3D collaborative filtering method is proposed based on block-energy normalization.Key steps of the method are as follows.Firstly,the block data were processed using soft/hard threshold filtering and block-energy normalization in the 2D transformation (1D transformation in the row (column) direction was performed first,followed by 1D transformation in the column (row) direction).Then block-matching and grouping in the 2D transformation domain,and 1D transformation in the block sequence direction were carried out,thereby achieving the 3D transformation of the block data.Soft/hard threshold filtering in the 3D transformation domain was performed,followed by 3D inverse transformation of the block data,and block aggregation with preservation of the energy relationship between block data.Theoretical analysis,model data testing,and application to actual data all showed that the proposed method can eliminate the influence of the energy difference between blocks,thereby improving the matching accuracy and computational efficiency across similar blocks,and ultimately improving the denoising effect,which can be widely used in denoising.

国家自然科学基金项目(41874168)、国家重点研发计划(2017YFB0202904)和四川省杰出青年科技计划(2019JDJQ0053)共同资助。