本文探讨了神经网络用于图像压缩的方法，提出了一种基于边缘信息检测的BP神经网络压缩方法:按图像各部分像素特征将图像分为平滑块、纹理块和边缘块，不同的分类块采用不同的隐含层数。实验结果证明本算法在在保证重建图像丰富细节的同时，能大大提高图像的压缩比。-This paper discusses the neural network for image compression method, edge information is presented based on BP neural network detection of compression method: According to the various parts of the image pixels will be characterized by a smooth image is divided into blocks, the texture blocks and edge blocks, different types of blocks implied different layers. Experimental results show that the algorithm in detail in ensuring the reconstruction of the image-rich at the same time, can greatly improve the image compression ratio.