The Research Of Color Images Recognition Based On Graying Code With Chromaticity Information

Image is one of the most important human information sources, and vision is the most significant perception system as well. The spatial characteristics of two-dimensional images result in image involving vast quantities of information, hence it is expected that computers are able to simulate the ability of human to understand and comprehend images as well as analysis and recognize the specific images information effectively. Human vision, besides its properties of size-invariance, location-independence. What's more, it is human interpretation to image containing knowledge and experiences. Artificial Neural Networks(ANN) are interconnected widely by large amounts of processing unit(neuron) ,which is to abstract, simplify and simulate human brain and posses natural characteristic of storing and applying prior knowledge.We propose a novel encoding approach that copes with transferring color to gray-scale images, that is an encoding rule of color image gray-processing while retaining chromaticity information. We present function modules of our image recognition system and investigate fundamental theory and key technique of ANN-based color image recognition. In order to remove interference information, we pre-process the images, including compression, gray-processing and fuzziness, and then we adopt Hamming Net to recognize the training samples.Based on the proposed approach, we take advantage of Object-Oriented Analysis and Design methodology to establish an image recognition system. The whole experiments are performed under Visual Studio.Net 2008 by C++. The results indicate that our method offers many benefits in smaller storage and quicker recognition rate while retaining chromaticity information. In short, our method is effective as well as efficient.