基于灰度DAG熵最大化量化分辨率医学图像增强

Gray level DAG maximum entropy based on quantization resolution for Medical image tone enhancement

摘要:

为提高医学图像增强的清晰度和对比度，并提高计算效率，提出一种基于灰度有向无环图(Directed acyclic graph，DAG)熵最大化量化分辨率医学图像色调增强算法。首先，采用简单的分段自回归（Piecewise autoregressive，PAR）模型进行图像目标恢复，并考虑到模数转换的误差利用全最小二乘算法进行PAR模型参数估计，获得高分辨率图像恢复直方图模型；其次，针对可能存在的对比度过低问题，将上述获得的最小二乘算法约束优化问题，建模为DAG中的最大权重路径问题，构建了色调保持最大熵图像增强过程约束优化模型，并通过DAG 图Monge定理特性实现计算复杂度的降低；通过上述两个步骤，实现了医学图像增强过程中图像细节和对比度的同步增强，仿真实验显示所提算法可提供更为有效的医学图像增强效果。

Abstract:

In order to improve the medical image sharpness and contrast, and improve the computational efficiency, we proposed the gray level DAG maximum entropy based on quantization resolution for Medical image tone enhancement. Firstly, we used a simple piecewise autoregressive (Piecewise autoregressive PAR) image target model for recovery, and taked into account the error of analog to digital conversion to use least squares algorithm to estimate PAR model parameter, which obtain high resolution image histogram restoration model; Secondly, aiming at the problem of low contrast may exist, the least squares algorithm for constrained optimization problems was modeled in DAG, which constructed a hue preserving constraint optimization model of maximum entropy image enhancement, and the characteristics of the DAG figure Monge theorem was used to reduce the computational complexity; Through the above two steps, the image details and contrast enhancement in the process of medical image enhancement are realized. The simulation results show that the proposed algorithm can provide more effective medical image enhancement effect.