基于图像处理参数反馈自适应方法设计与应用

Design and application of adaptive feedback method based on image processing parameters

摘要:

自动图像处理程序的参数优化是一个耗时过程。对于具有高噪音和阴影复杂图像分析任务,手动调整参数无法产生较好的效果。为了多参数同时调优,提出了基于反馈的参数自适应模型来改进标准图像分割方法。通过前馈方式实现算法,对其调整参数来比较算法的性能,根据基准数据集对算法进行评估、比较来讨论图像阴影和噪声对分割和分类精度的影响。研究结果表明,在不同阴影等级的背景下,采用反馈参数自适应的图像分割算法比采用前馈算法的图像分割、分类效果更好。当仅存在抽象基准数据时,该方法执行自动图像分析的有效,同时,使用分级的数据集评估不同图像处理的鲁棒性,有利于最终用户进行图像处理。为自动化图像处理理论研究与实践提供借鉴。

Abstract:

Parameter optimization of automatic image processing program is a time-consuming process. For complex image analysis tasks with high noise and shadow, manual adjustment of parameters can not produce good results. In order to optimize multiple parameters simultaneously, a parameter adaptive model based on feedback is proposed to improve the standard image segmentation method. The performance of the algorithm is compared by adjusting its parameters. The algorithm is evaluated and compared according to the benchmark data set to discuss the influence of image shadow and noise on segmentation and classification accuracy. The results show that, in different shadow levels, the image segmentation and classification effect of adaptive feedback parameter is better than that of feedforward algorithm. When there are only abstract reference data, this method is effective in performing automatic image analysis. At the same time, hierarchical data sets are used to evaluate the robustness of different image processing, which is beneficial to the end user for image processing. It provides reference for the theoretical research and practice of automatic image processing.