Corner Detection Based on Improved Shearlet Transform

Image corners have important theoretical significance and application value in the fields of image processing and analysis, machine vision, etc., however, the existing multi-scale analysis corner detectors fail to make full use of the multi-scale and multi-direction structural information, which affects their detection accuracy and diagonal positioning ability. Based on the basic principle of shear wave transform, the limitations of multi-scale and multi-direction image decomposition are analyzed, and the scale function is improved to the Butterworth type with a relatively flat falling edge, and then a new non-zero integrable function is obtained for the auxiliary function, and then a new multi-scale and multi-directional image decomposition method based on improved shear wave is proposed. Experimental results show that the proposed method not only partially improves the small-scale texture effect obtained after multi-scale image decomposition, but also solves the problem of large difference and messy texture between the previous small-scale local image and the original image of the clipping wave, and the image reconstruction effect is also improved and has strong robustness. In the field of image corner detection, the algorithm can make full use of multi-scale and multi-direction structural information to improve the positioning accuracy of image corner detection, reduce the error detection and missed detection rate, and has good robustness. 图像角点在图像处理与分析、机器视觉等领域具有重要理论意义与应用价值，然而，现有的多尺度分析角点检测器未能充分利用多尺度和多方向的结构信息，这影响了它们的检测精度和对角点的定位能力。基于剪切波变换的基本原理，分析其在图像多尺度、多方向分解的局限性，将尺度函数改进为较为平坦下降沿的巴特沃斯型，之后针对辅助函数改进得到新的非零可积函数，进而提出新的基于改进剪切波的多尺度、多方向图像分解方法。实验结果表明该方法不仅在图像多尺度分解后得到的小尺度的纹理效果得到部分提升，解决了之前剪切波小尺度局部图像与原图像差异性大、纹理杂乱的问题，而且图像重构效果亦有所改进且具有较强的鲁棒性。在图像角点检测领域利用该算法可充分利用多尺度和多方向的结构信息进而提升图像角点检测的定位精度，降低错检、漏检率，而且具有较好的鲁棒性。