CAJ | 학술논문

在研究和分析传统方差法和最大类间方差法的基础上，提出了一种传统方差法和最大类间方差法相结合的分块处理分割算法。该算法首先求出整个指纹图像的方差，然后计算每一子块的方差。若方差小于整个图像的方差则用方差法对图像进行分割，否则用最大类间方差法进行分割。最后再对图像进行平滑处理。实验结果表明，相对于传统分割方法，该方法无需根据经验选取阈值，能较准确地分割出前景和背景，抗噪能力强，对高对比度和低对比度的图像均具有很好的分割效果。
재연구화분석전통방차법화최대류간방차법적기출상，제출료일충전통방차법화최대류간방차법상결합적분괴처리분할산법。해산법수선구출정개지문도상적방차，연후계산매일자괴적방차。약방차소우정개도상적방차칙용방차법대도상진행분할，부칙용최대류간방차법진행분할。최후재대도상진행평활처리。실험결과표명，상대우전통분할방법，해방법무수근거경험선취역치，능교준학지분할출전경화배경，항조능력강，대고대비도화저대비도적도상균구유흔호적분할효과。
After investigating and analyzing the traditional variance method and the maximum between-cluster variance, this paper presents a new approach to segment the fingerprint images based on the maximum between-cluster variance and sub-block processing of the traditional variance segmenting. The method firstly calculates the variance of the fingerprint image and each subblock. Traditional variance is used to segment the image if the variance of each sub-block is less than the variance of the entire image. Otherwise, the OTSU method is used and finally smooth the fingerprint image. The experimental results indicate that this method does not need to choose the threshold with the experience compared with the traditional methods, and the can accurately separate the prospects from context. This method performs well on both of the fingerprint images with high and low contrast. The segment algorithm not only provides good segment result but also is robust to reduce noise.