分数阶微分的婴幼儿脑MR图像增强算法

发布时间：2018-04-18 10:13

本文选题：图像增强 + 自适应　；参考：《中国图象图形学报》2016年12期

【摘要】：目的由于传统的分数阶微分算法本质是提高相邻像素点的灰度差,达到增强对比度的作用,但是同时会放大和产生噪声,这容易使婴幼儿脑MR图像的增强效果有限或过增强。为了解决上述问题,提出一种融合非局部均值信息的自适应分数阶微分的婴幼儿脑MR图像增强算法。方法用平均梯度和大津算法自适应确定分数阶阶数,融合纹理粗糙度确定初始的分数阶阶数。为了进一步抵制噪声等干扰,利用更大邻域的纹理信息,融入非局部思想确定分数阶微分的阶数。最后用最终的分数阶阶数对图像进行滤波,得到最终的增强图像。结果实验通过信息熵、平均梯度和空间频率指标统计结果证明本文算法具有优越的图像增强性能。信息熵指标能够高出对比算法0.2%12%,平均梯度指标能够高出对比算法5%59%,空间频率指标能够高出对比算法6%59%。结论本文算法可以在增强纹理细节及抑制分数阶微分引入噪声方面都取得较好的效果。本文算法也适用于普通的模糊图像,具有良好的应用背景。
[Abstract]:Objective because the essence of the traditional fractional differential algorithm is to increase the gray difference of adjacent pixels and enhance the contrast, but it can enlarge and produce noise at the same time, it is easy to make the enhancement effect of infantile brain Mr image limited or too enhanced.In order to solve the above problem, an adaptive fractional differential image enhancement algorithm for infantile brain is proposed.Methods the fractional order is determined adaptively by the mean gradient and the Otsu algorithm, and the initial fractional order is determined by the fusion texture roughness.In order to further resist noise and other interference, the order of fractional differential is determined by using the texture information of larger neighborhood and incorporating the nonlocal idea.Finally, the final enhancement image is obtained by filtering the image with the final fractional order.Results the results of information entropy, average gradient and spatial frequency statistics show that the proposed algorithm has excellent image enhancement performance.The information entropy index can be higher than the contrast algorithm 0.2 points, the average gradient index can be higher than the contrast algorithm 5Y, and the spatial frequency index can be higher than the contrast algorithm 6Y.Conclusion the proposed algorithm can enhance texture details and suppress fractional differential noise.This algorithm is also suitable for common blurred images, and has a good application background.
【作者单位】：东北大学信息科学与工程学院;教育部医学影像计算重点实验室;
【基金】：国家自然科学基金项目(61370152,61402298) 中央高校基本科研业务费专项资金(N130204003)~~
【分类号】：TP391.41

【相似文献】