提出一种图像的小波系数预测编码方法二在该方法中，原始图像被分解为一个小尺寸的低频图像和一系列不同方向的多分辨率细节子带图像（小波系数图像），由于不同分辨率的小波系数图像具有相似性，故可用低分辨率小波系数来预测同一方向相邻较高分辨率的小波系数．采用Ｂ—Ｐ神经网络作为非线性预测器有效地实现了这种预测，结果表明，利用这种编码算法可获得高图像压缩比． In this method, the original image is decomposed into a small-size low-frequency image and a series of multi-resolution detail subband images (wavelet coefficient images) of different directions. Because of different resolutions Of the wavelet coefficients have similar image, so low resolution wavelet coefficients can be used to predict the higher resolution adjacent wavelet coefficients in the same direction. This prediction is effectively implemented using the B-P neural network as a nonlinear predictor, and the results show that a high image compression ratio can be obtained using this encoding algorithm.