基于神经网络的分形图像编码边缘信息提取

IMAGE CONTOUR EXTRACTION USING NEURAL NETWORK-BASED FRACTAL CODING

利用并行神经网络实现分形图像编码时 ,提出一种限定父块选取的编码方法 .该方法保留了结构简单、计算量小等特点 ,获得的分形码流不仅可从任意图像开始 ,通过多次迭代重建图像 ,而且还可以从均匀灰阶图像 (全黑图像除外 )出发 ,迭代一次获得精确的图像边缘 .模拟结果表明该方法是有效的 .

A coding method was presented to fix the selection of domain blocks when implementing fractal image coding with parallel neural network technologies. The obtained fractal code can be used to get the fine image contour after one iteration from any uniform gray level images (except all black), as well as to reconstruct the original image after many times of iteration from any images. At the same time, this method still keeps the features of simple structure and low computation complextiy. The simulation results show that this method is effective.