一种基于多光谱波段选择的白细胞图像分类法

基于多光谱波段的优化理论，在白细胞图像各波段问相关性分析的基础上，对传统最大似然分类法进行改进，提出一种对白细胞图像选择最有效部分波段进行识别的分类方法．该方法通过贝叶斯理论与其他先验知识进行融合，提高了分类准确度，克服了传统最大似然分类法在图像识别过程中具有的数据量庞大、计算程度繁冗和识别速度慢等缺点．实验结果表明，在保证分类精度基本不变的条件下使计算效率提高了2倍以上，为今后图像特征提取和图像分割奠定良好的基础。

A Microscopic Cell Image Recognize Method Based on Band Selection

Based on the optimum theory of multi-light spectrum bands and the analysis of the inter-(relation) of various bands of the pictures of white cells, a new classification method-a great improvement on the traditional maximum likelihood classification-has been suggested which can classify the most effective parts of the bands of the white cell pictures. By combining Bayesian principles and other priori knowledge, the method has improved the degree of accuracy of classification and overcome shortcomings of immense data quantity, complexity of calculation and slow speed of recognition which exist in traditional maximum likelihood classification in recognizing pictures. The experiment results show that the calculation efficiency has increased nearly twice fold while ensuring there will not be a main change in the degree of accuracy in classification, which has laid a good foundation for future extraction of the characteristics as well as (breaking)-up of pictures.