Measurement Model for Medical Image Feature Matrix Similarity Based on CNN

The original similarity measurement model is easy to ignore the processing of image details, resulting in poor accuracy measurement. In paper, we propose a for medical feature matrix based on convolutional neural network (CNN). First, Gaussian convolution kernel used obtain global and local data images, corresponding set formed. Second, layer CNN introduced, obtained by layer. Finally, constructed. results show that effect this better when test process divided into three parts: global, local, detail. highest error rate proposed algorithm only about 0.21, which takes less time, overall fitting degree can reach 91%. Compared with traditional methods, higher use better.