Scholarlink

<p>The expansion of multimedia applications, fueled by progress in internet technologies and communication infrastructure, has underscored the need for effective image compression techniques to manage the growing demands for storage and bandwidth. Lossless image compression, crucial for high-cost image acquisition scenarios and applications requiring exact data preservation, presents significant challenges in balancing compression efficiency with data fidelity. This paper introduces a novel approach to lossless image compression that enhances performance through advanced statistical modeling and encoding techniques. Specifically, our method integrates Finite Mixture Models (FMMs) and Bayesian modeling with spatial domain encoding to overcome traditional limitations. Unlike conventional methods that operate in the prediction error domain, our approach encodes pixel values directly in the spatial domain, mitigating the zero-frequency problem and improving compression efficiency. By employing a Laplace distribution to model prediction errors and utilizing FMMs to estimate the probability density function (PDF) of image data within a Bayesian framework, we achieve superior compression rates compared to leading standards such as JPEG-2000, JPEG-LS, and JPEG-XL. Our experimental results demonstrate an average compression rate of 3.96 bits per pixel (bpp) with our method, outperforming JPEG-2000, JPEG-LS, and JPEG-XL by <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(18\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>18</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation>, <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(14.5\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>14.5</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation>, and <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(6.8\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>6.8</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation>, respectively.</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

Enhancing lossless image compression with Bayesian mixture models and spatial domain encoding

  • 期刊论文

  • 发表日期：

    2025年12月13日

查看全文
  • Sonia Zouari,
  • Sarah Abu Ghazalah,
  • Atef Masmoudi,
  • Afif Masmoudi

摘要

The expansion of multimedia applications, fueled by progress in internet technologies and communication infrastructure, has underscored the need for effective image compression techniques to manage the growing demands for storage and bandwidth. Lossless image compression, crucial for high-cost image acquisition scenarios and applications requiring exact data preservation, presents significant challenges in balancing compression efficiency with data fidelity. This paper introduces a novel approach to lossless image compression that enhances performance through advanced statistical modeling and encoding techniques. Specifically, our method integrates Finite Mixture Models (FMMs) and Bayesian modeling with spatial domain encoding to overcome traditional limitations. Unlike conventional methods that operate in the prediction error domain, our approach encodes pixel values directly in the spatial domain, mitigating the zero-frequency problem and improving compression efficiency. By employing a Laplace distribution to model prediction errors and utilizing FMMs to estimate the probability density function (PDF) of image data within a Bayesian framework, we achieve superior compression rates compared to leading standards such as JPEG-2000, JPEG-LS, and JPEG-XL. Our experimental results demonstrate an average compression rate of 3.96 bits per pixel (bpp) with our method, outperforming JPEG-2000, JPEG-LS, and JPEG-XL by \(18\%\) 18 % , \(14.5\%\) 14.5 % , and \(6.8\%\) 6.8 % , respectively.