论文部分内容阅读
首先将煤矿工业电视图象分为相对静止区和相对运动区两部分．其次对相对运动区通过小波变换将图象信号分解为若干不同空间分辨率、频率特性和方向特性的子图象信号，然后利用基于ＢＰ算法的三层前向神经网络构造的非线性预测器，进行压缩编码传输，最后和单独处理的相对静止区图象合成输出，收到了较好的效果。
First of all, the coal industry television image is divided into two parts: the relative quiescent zone and the relative movement zone. Secondly, the relative motion region is decomposed into several sub-image signals with different spatial resolution, frequency characteristic and direction characteristic by wavelet transform. Then, a nonlinear predictor constructed by three-layer feedforward neural network based on BP algorithm is used, Compression encoding and transmission, and finally processed separately and relatively still image synthesis output, received good results.