小尺度图像细节层的高频分量提取仿真
|
摘要
目前视觉领域针对图像处理没有相对完善的方案,缺少有效提取图像的高频分量方法,因此采用当前方法对小尺度图像细节层的高频分量进行提取时,存在分量提取效果不佳、图像细节信息不显著以及视觉效果不好等问题,提出一种基于人工神经网络模型的小尺度图像细节层高频分量提取方法。首先对待处理的小尺度图像进行双边滤波去噪,对平滑图像进行离散小波变换,提取出图像细节层中水平细节分量、垂直细节分量、对角细节分量;然后采用人工神经网络模型对提取出的三个细节层分量进行增强处理,进一步体现细节层的高频分量,使提取的高频分量更为平滑。仿真实验证明,所提方法对小尺度图像的高频分量提取精度较高、所需时间较短。
At present, the method for effectively extracting high-frequency component of image in the field of vision is lacked. In current methods, the component extraction effect is poor and the image detail information is not significant. Therefore, a method for extracting high-frequency components in detail layer of small-scale image based on artificial neural network model was presented. Firstly, the bilateral filtering de-noising was performed on the small-scale image to be processed. Secondly, the discrete wavelet transform was applied to the smoothed image, so as to extract the horizontal detail component, the vertical detail component and the diagonal detail component in image detail layer. Then, the artificial neural network model was used to enhance the components in three detail layer and thus to further reflect the high frequency components in detail layer. Thus, the extracted high frequency components could be smoother. Simulation proves that the proposed method has high precision and low time consumption for extracting the high-frequency component in small-scale image.
At present, the method for effectively extracting high-frequency component of image in the field of vision is lacked. In current methods, the component extraction effect is poor and the image detail information is not significant. Therefore, a method for extracting high-frequency components in detail layer of small-scale image based on artificial neural network model was presented. Firstly, the bilateral filtering de-noising was performed on the small-scale image to be processed. Secondly, the discrete wavelet transform was applied to the smoothed image, so as to extract the horizontal detail component, the vertical detail component and the diagonal detail component in image detail layer. Then, the artificial neural network model was used to enhance the components in three detail layer and thus to further reflect the high frequency components in detail layer. Thus, the extracted high frequency components could be smoother. Simulation proves that the proposed method has high precision and low time consumption for extracting the high-frequency component in small-scale image.
引文
[1] 韦瑞峰,赵荣普,徐肖庆,等.基于直方图的红外图像细节增强算法研究[J].红外技术,2016,38(6):472-475.
[2] 田馨,廖明生.精确提取InSAR时间去相关分量的方法[J].红外与毫米波学报,2016,35(4):454-461.
[3] 于来行,冯林,张晶,et al.自适应融合目标和背景的图像特征提取方法[J].计算机辅助设计与图形学学报,2016,28(8):1250-1259.
[4] 曾接贤,祝小超,符祥.一种改进的复杂图像线特征提取方法[J].中国图象图形学报,2018,15(12):1748-1754.
[5] 张元峰,郝世勇,司剑飞,等.基于传输矩阵的电力变压器高频网络参数计算[J].电子设计工程,2018,26(3):84-88.
[6] 于来行,冯林,张晶,等.自适应融合目标和背景的图像特征提取方法[J].计算机辅助设计与图形学学报,2016,28(8):1250-1259.
[7] 杨静,李争.一种基于双边滤波的红外图像细节增强方法[J].激光与红外,2016,46(4):507-511.
[8] 施丽红,刘刚.基于改进蜂群优化的图像分割算法[J].电视技术,2016,40(2):37-44.
[9] 卢献健,晏红波,梁月吉.小波分解层数及分量组合对滑坡预测的影响[J].桂林理工大学学报,2016,36(2):304-309.
[10] 郝志成,吴川,杨航,等.基于双边纹理滤波的图像细节增强方法[J].中国光学,2016,9(4):423-431.
[11] 巨刚,袁亮,刘小月,等.多算法融合的自适应图像增强方法[J].光子学报,2016,45(12):136-144.
[12] 谢洪森,周鹏,栾宝宽,等.一种海杂波纹理分量提取方法研究[J].雷达科学与技术,2017,15(3):251-254.
[2] 田馨,廖明生.精确提取InSAR时间去相关分量的方法[J].红外与毫米波学报,2016,35(4):454-461.
[3] 于来行,冯林,张晶,et al.自适应融合目标和背景的图像特征提取方法[J].计算机辅助设计与图形学学报,2016,28(8):1250-1259.
[4] 曾接贤,祝小超,符祥.一种改进的复杂图像线特征提取方法[J].中国图象图形学报,2018,15(12):1748-1754.
[5] 张元峰,郝世勇,司剑飞,等.基于传输矩阵的电力变压器高频网络参数计算[J].电子设计工程,2018,26(3):84-88.
[6] 于来行,冯林,张晶,等.自适应融合目标和背景的图像特征提取方法[J].计算机辅助设计与图形学学报,2016,28(8):1250-1259.
[7] 杨静,李争.一种基于双边滤波的红外图像细节增强方法[J].激光与红外,2016,46(4):507-511.
[8] 施丽红,刘刚.基于改进蜂群优化的图像分割算法[J].电视技术,2016,40(2):37-44.
[9] 卢献健,晏红波,梁月吉.小波分解层数及分量组合对滑坡预测的影响[J].桂林理工大学学报,2016,36(2):304-309.
[10] 郝志成,吴川,杨航,等.基于双边纹理滤波的图像细节增强方法[J].中国光学,2016,9(4):423-431.
[11] 巨刚,袁亮,刘小月,等.多算法融合的自适应图像增强方法[J].光子学报,2016,45(12):136-144.
[12] 谢洪森,周鹏,栾宝宽,等.一种海杂波纹理分量提取方法研究[J].雷达科学与技术,2017,15(3):251-254.