Scholarlink

<p>Modern steganography must simultaneously ensure high embedding capacity, visual fidelity of the covering medium, and resistance to detection, ideally while improving perceptual quality. This paper introduces two spatial-domain steganographic methods that achieve high embedding capacity for least-significant-bit (LSB)-based approaches while producing stego images with improved perceptual quality relative to the cover images through integrated contrast enhancement and global color balancing. The first technique operates on grayscale and color images, and the second is optimized specifically for RGB images. Experiments on three standard benchmark datasets (KODAK, SIPI, and INRIA Holidays) produce embedding rates of 1 bit per pixel for grayscale images and 3 bits per pixel for RGB images. Under high-capacity embedding conditions, the first method achieves a peak signal-to-noise ratio of 34.58 decibels with negligible color distortion (CIEDE2000 metric: 1.79), and the second method achieves a peak signal-to-noise ratio of 34.17 decibels with a BRISQUE of 14.92. Importantly, the relative contrast and edge enhancement metrics consistently exceed 0.5, objectively confirming improved perceptual quality. A comparative ablation analysis shows that the enhancement module consistently reduces detection accuracy for both the StegExpose and SRNet evaluators on the INRIA Holidays and BOSSbase datasets. These results demonstrate that a consistent balance between high embedding capacity and visual quality can be maintained while also achieving perceptual enhancement and statistical security, addressing a practical limitation in contemporary data-hiding 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

A unified framework for data hiding and image enhancement in grayscale and RGB domains

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月16日

查看全文
  • Anggoro Suryo Pramudyo,
  • Faizal Makhrus,
  • Reza Pulungan

摘要

Modern steganography must simultaneously ensure high embedding capacity, visual fidelity of the covering medium, and resistance to detection, ideally while improving perceptual quality. This paper introduces two spatial-domain steganographic methods that achieve high embedding capacity for least-significant-bit (LSB)-based approaches while producing stego images with improved perceptual quality relative to the cover images through integrated contrast enhancement and global color balancing. The first technique operates on grayscale and color images, and the second is optimized specifically for RGB images. Experiments on three standard benchmark datasets (KODAK, SIPI, and INRIA Holidays) produce embedding rates of 1 bit per pixel for grayscale images and 3 bits per pixel for RGB images. Under high-capacity embedding conditions, the first method achieves a peak signal-to-noise ratio of 34.58 decibels with negligible color distortion (CIEDE2000 metric: 1.79), and the second method achieves a peak signal-to-noise ratio of 34.17 decibels with a BRISQUE of 14.92. Importantly, the relative contrast and edge enhancement metrics consistently exceed 0.5, objectively confirming improved perceptual quality. A comparative ablation analysis shows that the enhancement module consistently reduces detection accuracy for both the StegExpose and SRNet evaluators on the INRIA Holidays and BOSSbase datasets. These results demonstrate that a consistent balance between high embedding capacity and visual quality can be maintained while also achieving perceptual enhancement and statistical security, addressing a practical limitation in contemporary data-hiding systems.