相位一致性加权的引导图像滤波去噪算法

针对传统引导图像滤波算法采用固定的正则化参数规整化因子造成光晕伪影现象的问题,提出一种相位一致性加权自适应规整化因子的改进引导图像滤波算法.因图像不同区域的纹理特性不同,利用相位一致性能够获取图像丰富的特征信息和精确的特征定位,对原有的规整化因子在不同区域通过惩罚自适应调整规整化因子,保证了图像降噪效果同时进一步保持边缘清晰.实验采用结构相似性因子(SSIM)与PSNR作为算法的定量评估指标.结果表明,采用相位一致性作为加权值规整化因子,有效地避免了传统引导图像滤波的光晕伪影现象,相对于传统引导图像滤波算法,降噪性能有一定提高,还能够很好地保持边缘细节信息.

Weighted Guided Image Filtering Algorithm Using Phase Congruency for Image Denoising

In order to solve the problem that traditional guided image filtering algorithm using a fixed regularization parameter regularization factor caused halo artifacts,a weighted guided image filtering algorithm using phase congruency was proposed. As the textural properties vary in different regions of the image,the phase coherence can obtain richly image features and precisely targeting information. Adaptive adjustment penalty factor in different areas using monogenic phase congruency was to ensure the image noise effect and further maintain sharp edges. Structural similarity( SSIM)and PSNR were used to evaluate the quality of images in experiments. The experimental results showed that the monogenic phase coherence,as adaptive penalty factor,can effectively avoid the halo artifact phenomenon of the traditional guided image filtering. Denoising performance is not only improved,but also good edge details are retained.