본문 바로가기
  • Home

Adaptive Search Patch Selection Preprocessing in NLM Filtering Based on DCT and FFT Frequency Analysis

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(10), pp.113~120
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : September 1, 2025
  • Accepted : September 29, 2025
  • Published : October 31, 2025

Jiyun Byun 1 Hyojin Lee 1 Hongrae Lee 1

1연세대학교

Accredited

ABSTRACT

With the increasing use of visual media, the importance of image processing has grown, and effective denoising techniques are essential as noise directly affects image analysis. The Non-Local Means (NLM) filter, which removes noise by searching for similar regions, is widely used but limited because a fixed search patch size does not account for regional characteristics. This paper proposes a preprocessing method that transforms images into the frequency domain and adaptively adjusts the search patch size of the NLM filter according to frequency-based region separation. When applied to the NLM filter, the proposed preprocessing technique improved the PSNR by an average of 2.50 dB and up to 8.24 dB, confirming its effectiveness in enhancing denoising performance.

Citation status

* References for papers published after 2024 are currently being built.