Abstract: A Discrete Cosine Transform（DCT） based image compression algorithm is proposed in this paper.The compressed image data array is limited to a certain length by setting an adaptive quantization truncating threshold.A fast DCT algorithm is implementation，and an efficient non-zero quantized data storage method is used to simplify the computation.The traditional entropy coding is not put into use.In the decoding side，the concept of adaptive-neighborhood statistics is applied to eliminate the blocking artifact which is resulted in quantization.The proposed method makes further improvement for image visual quality.Furthermore，it plays a good performance for real-time online image processing.In the end，the proposed DCT based method is applied into a remote close-loop vision control system of a 2-DOF robot arm.

Key words: image compression, Discrete Cosine Transform（DCT）, blocking artifact, postfiltering, adaptive-neighborhood, Peak Signal to Noise Ratio（PSNR）

摘要： 利用基于二维离散余弦变换（DCT）的方法进行图像压缩，通过设置自适应量化截断阈值来控制图像压缩数组的大小。为了加快运算速度，在图像压缩部分利用快速DCT算法；并直接对量化后的非零系数进行存储，而不采用熵编码的方式。在解码端，针对量化产生的块效应，利用基于自适应邻域的思想进行后处理。所提方法进一步提高了图像的峰值信噪比和主观视觉质量，并且达到了实时在线处理的要求。文中所提基于DCT的图像压缩方法在一个基于视觉位置反馈的二自由度机械臂远程控制系统中得到实际应用。

关键词: 图像压缩, 离散余弦变换, 块效应, 后滤波, 自适应邻域, 峰值信噪比