该文从图像局部结构特性出发,提出一种基于图像局部结构信息的空间汇集算法.首先对原始图像进行高斯模糊,提取局部结构信息并进行非零处理,进而形成汇集权重图.通过对PSNR和SSIM的加权实验,验证了该汇集算法能够明显提高图像质量评价算法性能,并且对各种失真类型具有较好的适应性.%A spatial pooling strategies (SPS) for image quality assessment (IQA) is proposed, which obtain a weight map from local structural information (LSI) of the reference image. The reference image is first smoothed with a Gaussian kernel, and the LSI is extracted from the smoothed image. A weight map is derived from the LSI using a non-zero procedure. PSNR and SSIM obtained in the experiments show effectiveness of the proposed method for a wide range of image distortion types.
展开▼