面向星载应用的空间域四叉树层次化图像压缩算法

Image Compression Algorithm Based on the Quarter-tree for Satellite Remote Sensing Image Coding

为满足卫星应用领域对高分辨率遥感图像实时传输与存储的要求,针对该类图像特征,提出基于空间域四叉树数据结构的图像数据亚采样与多模式自适应预测编码相结合的压缩算法。该算法使用四叉树数据结构与三种预测编码模式,针对图像块纹理差别,自适应地选取相应的亚采样与量化编码方式,通过采样方式与量化方式的变化,达到保存高分辨率遥感图像细节与小目标绝不丢失的要求,实现了对遥感图像中目标边缘和变化剧烈的细节的高保真效果,同JPEG相比,恢复图像峰值信噪更高、图像纹理细节保持能力更强。利用具有可编程序门阵列器件物理实现上述算法,获得高速图像压缩专用芯片,该芯片数据处理速度达到288Mbit/s,功耗低于1W。

In order to resolve data rates in data transmission compression algorithm of combini the contradiction between the need of high image quality and low and storage in the fields of satellite remote sensing, a new ng spatial quarter-tree data architecture with m.ulti-mode adaptive quantization technique is proposed. The algorithm is characterized with higher PSNR of the reconstructed image and lower computation complexity than JPEG, and the image detail preserving capability of the algorithm is better than that of JPEG. The ASIC realizing the algorithm is designed and manufactured using Field Programmable Gate Arrays. The peak processing speed of the ASIC can reach 288Mbit/s and its power consumption is lower than 1W.