基于多特征发现的生理切片骨小梁识别技术研究

目标区域计算机自动识别是骨组织形态计量学的发展趋势。研究图像多特征信息,识别苏木素尹红染色病理切片中的骨小梁。分析骨组织病理切片HSL颜色空间通道的直方图特征,进行基于颜色通道的阈值分割,采用Lazy Snapping算法获得骨小梁深层次特征信息,改进扫描种子填充算法,使其融合目标区域多特征信息。实验结果表明改进的扫描种子填充算法将复杂图像骨小梁提取时间从常规软件的10多分钟提速到257ms。改进的扫描线算法,能快速获得稳健的目标区域。

Study on Physiological Slice Trabecular Recognition Based on Multi-feature-found

Target recognition is a development trend of bone histomorphometry. Study on the characteristics of image information, it is necessary to discriminate trabecular bone of bone tissue sliced by yin hung-hematoxylin. Analysis bone tissue image histogram channel characteristics of HSL color space, based on the image segmentation threshold of color channels, by Lazy Snapping method, it will obtain the image features of trabecular bone in-depth information, and it will improve scanning seed filling algorithm to extract multi feature information by fusion target area. The results show that extract trabecular bone by analysis image multi-feature let repetitive operation improve, improved scanning seed filling algorithm let complex image such as trabecular extraction speed from ten minutes to 257ms. Improved scan line algorithm can get the robust target area more quickly.