Scholarlink

<p>This paper introduces an innovative reconfigurable AI-enabled vectored median filter designed for FPGA-based multi-image processing systems. The central focus is on mitigating image quality degradation caused by mixed impulsive noise during the transmission and reception of images in various applications, diverging from the conventional emphasis on smart vision systems. In our proposed integrated system, we integrate an RGB-Y filter with an AI-enabled impulse&#xa0;detection and impulse de-noising-based image filter. This pioneering approach effectively suppresses noise in multi-color images while preserving essential edge details crucial for diverse smart vision systems. The algorithm introduces a novel technique of substituting noisy pixels with processed central values within the image filtering window, ensuring fidelity to the original pixel—an imperative consideration for applications beyond traditional smart imaging. To address elevated noise densities encountered in a broad spectrum of applications, our methodology adopts a hybrid sorting approach for median filtering and impulse processing. This ensures robust noise reduction without imposing excessive computational complexity. The AI-enabled vectored median filter system achieves a noteworthy reduction in dynamic power consumption, showcasing a remarkable 46% decrease in power consumption and an 82% reduction in area when compared to existing systems. This holds significant advantages in meeting resource and power-aware constraints across various smart vision systems. A comprehensive performance evaluation, including metrics such as PSNR (Peak Signal-to-Noise Ratio), MSE (Mean Squared Error), IEF (Image Enhancement Factor), and SSIM (Structural Similarity Index), validates the effectiveness of the filter in enhancing image quality—a pivotal factor for a diverse range of multi-image processing systems.</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

Reconfigurable AI-enabled vectored median filter for real-time image denoising and edge preservation in FPGA-based smart imaging systems

  • 期刊论文

  • 发表日期：

    2025年2月11日

查看全文
  • Nanduri Sambamurthy,
  • Maddu Kamaraju

摘要

This paper introduces an innovative reconfigurable AI-enabled vectored median filter designed for FPGA-based multi-image processing systems. The central focus is on mitigating image quality degradation caused by mixed impulsive noise during the transmission and reception of images in various applications, diverging from the conventional emphasis on smart vision systems. In our proposed integrated system, we integrate an RGB-Y filter with an AI-enabled impulse detection and impulse de-noising-based image filter. This pioneering approach effectively suppresses noise in multi-color images while preserving essential edge details crucial for diverse smart vision systems. The algorithm introduces a novel technique of substituting noisy pixels with processed central values within the image filtering window, ensuring fidelity to the original pixel—an imperative consideration for applications beyond traditional smart imaging. To address elevated noise densities encountered in a broad spectrum of applications, our methodology adopts a hybrid sorting approach for median filtering and impulse processing. This ensures robust noise reduction without imposing excessive computational complexity. The AI-enabled vectored median filter system achieves a noteworthy reduction in dynamic power consumption, showcasing a remarkable 46% decrease in power consumption and an 82% reduction in area when compared to existing systems. This holds significant advantages in meeting resource and power-aware constraints across various smart vision systems. A comprehensive performance evaluation, including metrics such as PSNR (Peak Signal-to-Noise Ratio), MSE (Mean Squared Error), IEF (Image Enhancement Factor), and SSIM (Structural Similarity Index), validates the effectiveness of the filter in enhancing image quality—a pivotal factor for a diverse range of multi-image processing systems.