The Arithmetic And Its Application Of Image Compression And Classification Based On Artificial Neural Networks

Artificial neural networks is a nonlinear information-processing system which simulating the human brain's structure and function, it can be used to abstract and simulate the human brain's basic characteristics using mathematic methods. Because of its parallel and distributing specialties, artificial neural network has great advantages on dealing with the information of visual thinking voice and image recognition,associative memory. The image compressing and recognizing technology based on artificial neural networks has been the researching hotspot, and various net structures come up.In this thesis, the arithmetic of digital image compression and classification using neural networks technology has been analyzed ,improved and applied. To begin with, an improved vector quantization coding method has been put forward for digital image's vector quantization coding. It rearranges the code vectors corresponding with the code book, which makes the image compressed arithmetic more reasonable in associative memory network.Secondly, for the sake of decreasing the bandwidth when transmitting the signals and getting better image when transmission error occurred, the digital image has been pretreated. The associative memory network arithmetic has been improved by introducing extended vector quantization coding method, and then, the digital image has been coded and compressed .The compressing process has been simulated by computer.At last, at the instance of on microprocessor and high timing demand,the digital image has been compressed by BP network, and then the compressed data has been classified by SOFM network. Experimental results show that it uses less storage and computing time in comparing with using initial image data with large dimensions.