Summary: | 碩士 === 亞洲大學 === 資訊工程學系碩士班 === 100 === The study presents an automatic target-preserving content-aware image resizing method for displaying image with different aspect ratios. The method computes a saliency map which describes the attractive and unattractive regions in an image by the graph-based visual saliency model. Then a thresholding technique is applied to detect important target regions on the saliency map. In our method, three types of weighting algorithms are proposed to calculate the height of each row and the width of each column outside the important regions. These three types of algorithms are the average weighting, the graph-based visual saliency weighting, and the distance transform weighting algorithms. Next an interpolation technique is used to resize the image to its target size. Experimental results show that the proposed image resizing method is capable of preserving important target without deformation.
Keywords:Image retargeting, Image resizing, Seam carving, Visual saliency, Distance transform, Interpolation
|