Algorithm-Architecture Matching for Signal and Image Processing
Algorithm-Architecture Matching for Signal and Image Processing
复制标题
信号和图像处理的算法架构匹配
DOI：
10.1007/978-90-481-9965-5_1
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
Chen X
中科院分区：
文献类型：
--
作者：
Chen X
本文提出了一种新颖的无损多模式带间图像压缩（LMMIC）方案及其硬件架构。我们的方法检测图像的局部特征，并使用不同的模式自适应地对具有不同特征的区域进行编码。运行模式用于同质区域，而三元模式和常规模式分别用于边缘和其他区域。在常规模式下，我们提出了一种简单的带移技术作为带间预测和基于梯度的切换策略来在带内或带间预测之间进行选择。我们还在运行模式和三元模式下启用带内和带间自适应。 LMMIC的优点是自适应地“分割”图像并使用合适的方法对不同区域进行编码。我们的方案的简单性使得硬件变得易于使用。实验结果表明，LMMIC 实现了卓越的压缩比，具有能够对任意数量的频段进行编码并轻松访问任意频段的优点。我们还描述了该方案的硬件架构。
This paper presents a novel Lossless Multi-Mode Interband image Compression (LMMIC) scheme and its hardware architecture. Our approach detects the local features of the image and uses different modes to encode regions with different features adaptively. Run-mode is used in homogeneous regions, while ternary-mode and regular-mode are used on edges and other regions, respectively. In regular mode, we propose a simple band shifting technique as interband prediction and a gradient-based switching strategy to select between intraband or interband prediction. We also enable intraband and interband adaptation in the run-mode and ternary-mode. The advantage of LMMIC is to adaptively “segment” the image and use suitable methods to encode different regions. The simplicity of our scheme enables the hardware amenability. Experimental results show that LMMIC achieves superior compression ratios, with the benefits of enabling encoding any number of bands and easy access to any band. We also describe the hardware architecture for this scheme.