CAJ | 학술논문

研究了一个3层BP网络的图像压缩方案，分别采用了不同的传输函数、不同的隐含层神经元数目和同一幅图像不同大小的分块方式，分析和总结BP网络压缩性能与各种网络参数之间的关系。结果表明：BP网络传输函数、隐含层神经元数目以及对原图像分块大小的不同，都会直接影响收敛时间和压缩性能。
연구료일개3층BP망락적도상압축방안，분별채용료불동적전수함수、불동적은함층신경원수목화동일폭도상불동대소적분괴방식，분석화총결BP망락압축성능여각충망락삼수지간적관계。결과표명：BP망락전수함수、은함층신경원수목이급대원도상분괴대소적불동，도회직접영향수렴시간화압축성능。
The image compression scheme of a 3-layer BP network was studied in this paper.The relationship between BP network compression performance and network parameters was analysed and summarized using different transmission functions,different numbers of the hidden layer neurons and different dividing ways of the same image. The results showed that the convergence time and compression properties was affected by BP network transmission function,the number of the hidden layer neurons and different dividing ways of the original image directly.