基于内容划分和傅里叶分析的图像缩放算法

Image resizing algorithm based on content partition and Fourier analysis

为满足图像缩放中保护重要区域和视觉连贯性的要求,算法通过对不同重要度的区域采用不同的采样率进行缩放。用显著区域、语义内容和结构信息的特征来检测图像中的重要区域,根据重要区域将原图像自适应地划分为多个子图,并根据傅里叶分析和视觉损失能量函数计算每个子图的采样率,对子图进行下采样得到最终的缩放图像。仿真实验表明,与Seam Carving等算法相比,该算法计算效率较高,而且对图像中的显著物体保护较好。

To satisfy the requirements of protecting important area and visual coherence while resizing images, different sampling rate is applied to different parts of the image according to their importance. The algorithm initially detected important area of input image containing salient region, semantic content and structural information, then partitioned input image into sub-images adaptively according to its impor- tance area, and then computed sample rate for every sub-image based on Fourier analysis and visual distortion energy function, finally sub- sampled sub-imaMges and obtained the target image. Experimental results show that the proposed algorithm provide effective perforrmnce and satisfactory results on preserving salient objects compared with other resizing techniques, such as seam carving