Scholarlink

<p>Scale-invariant feature transform (SIFT) is one of the common algorithms in image registration. The&#xa0;SIFT algorithm uses a Gaussian filter in the pre-processing step, which blurs the image. In this paper, we use a combination of complex nonlinear diffusion and phase congruency to highlight the boundaries and edges of the image in different directions and angles more clearly. The complex nonlinear diffusion shows a clearer image in the spatial domain and the Phase congruency (PC) uses the frequency domain information in the spatial domain. By using the complex diffusion process, a smooth image is obtained, which preserves the ramp edges in the image. PC is invariant against brightness and contrast variations and robust to nonlinear light radiation and speckle noise. In the stage of extracting key points, we propose an improved multi-scale Harris algorithm. The experiment results show that the matching accuracy has improved by about 20% after applying fast sample consensus (FSC) with respect to state-of-the-art other methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Registration of Remote Sensing Images by the Combination of Complex Nonlinear Diffusion and Phase Congruency Attributes

  • 期刊论文

  • 发表日期：

    2025年1月6日

查看全文
  • Fatemeh Khalili,
  • Farbod Razzazi,
  • S. Abolfazl Hosseini

摘要

Scale-invariant feature transform (SIFT) is one of the common algorithms in image registration. The SIFT algorithm uses a Gaussian filter in the pre-processing step, which blurs the image. In this paper, we use a combination of complex nonlinear diffusion and phase congruency to highlight the boundaries and edges of the image in different directions and angles more clearly. The complex nonlinear diffusion shows a clearer image in the spatial domain and the Phase congruency (PC) uses the frequency domain information in the spatial domain. By using the complex diffusion process, a smooth image is obtained, which preserves the ramp edges in the image. PC is invariant against brightness and contrast variations and robust to nonlinear light radiation and speckle noise. In the stage of extracting key points, we propose an improved multi-scale Harris algorithm. The experiment results show that the matching accuracy has improved by about 20% after applying fast sample consensus (FSC) with respect to state-of-the-art other methods.