基于数据和模型驱动的皮沟图像检测优化算法研究

RESEARCH ON OPTIMISED SKIN GROOVES DETECTION ALGORITHM BASED ON DATA DRIVEN AND MODEL DRIVEN CONTROL

皮肤沟壑检测是对图像的二维几何特征值计算并且进行客观量化的一个关键性步骤。基于数据驱动和模型驱动的综合考虑,提出一种混合控制策略来实现优化效果的皮沟检测。先通过数据驱动的控制实施图像相关增强等处理;然后,通过执行分水岭变换从而获得粗略的检测结果。在获得数据驱动处理结果以及视觉感知的先验知识的基础上,建立一个区域合并模型,以防止虚假皮沟的出现。通过实际案例验证了对不同粒度的皮肤图像,提出的方法检测精度及有效性均较高。

Skin grooves detection is a key step in calculating the two-dimensional geometric eigenvalue of image and making objective quantification.Based on the comprehensive consideration of data-driven and model-driven controls,in this paper we propose a hybrid control strategy to realise skin grooves detection with optimised effect.In the strategy,first the data-driven control is used to carry out the treatment of image correlated enhancement,etc.;then through applying the watershed transform,the rough detection result is obtained.In order to prevent false skin grooves,we establish a region merging model based on acquiring the data-driven processing results as well as the prior knowledge derived from visual perception.Actual cases have verified that for skin images with different granularities,the proposed method has higher detection precision and validity.