Scholarlink

<p>This paper proposes a fine-grained, predictive distortion model for transform-domain motion estimation, leveraging the Walsh–Hadamard Transform (WHT) to reduce computational complexity through dominant coefficient analysis. While existing WHT-based methods often suffer from nonlinear distortion accumulation, the proposed approach addresses this by applying WHT to spatially dispersed 2 × 2 sub-blocks. This design ensures a more uniform energy distribution and enables progressive partial distortion accumulation, facilitating efficient early termination. The near-linear growth of cumulative distortion allows reliable prediction of final distortion values from early-stage results. Experimental evaluations demonstrate that the proposed method achieves up to 55.39% computational savings, with a peak PSNR loss of only 0.38&#xa0;dB.</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

Predictive fine-grained transform-domain partial distortion fast matching algorithm for motion estimation

  • 期刊论文

  • 发表日期：

    2026年1月27日

查看全文
  • A. V. Paramkusam

摘要

This paper proposes a fine-grained, predictive distortion model for transform-domain motion estimation, leveraging the Walsh–Hadamard Transform (WHT) to reduce computational complexity through dominant coefficient analysis. While existing WHT-based methods often suffer from nonlinear distortion accumulation, the proposed approach addresses this by applying WHT to spatially dispersed 2 × 2 sub-blocks. This design ensures a more uniform energy distribution and enables progressive partial distortion accumulation, facilitating efficient early termination. The near-linear growth of cumulative distortion allows reliable prediction of final distortion values from early-stage results. Experimental evaluations demonstrate that the proposed method achieves up to 55.39% computational savings, with a peak PSNR loss of only 0.38 dB.