Hits:

Date of Publication:2014-01-01

Journal:计算机辅助设计与图形学学报

Affiliation of Author(s):机械工程学院

Issue:7

Page Number:1167-1175

ISSN No.:1003-9775

Abstract:In order to preserve the human and the salient objects within the image as well as avoid local region shrunk excessively, an image resizing algorithm is proposed based on the image energy map and strip dividing. Firstly, the gradient values of all pixels are computed and human region is detected and assigned a relatively large weight for an original image, thus the energy map of the image is constructed by combining the gradient values and the weight of the detected human region. Through iteration calculation, the image can be divided into several strips by integrating the similar energy-level columns together. Then, according to the average accumulated energy of each strip, the reduced size amount of a strip is determined in inverse proportion to its energy. The resized level of a strip is judged by setting a threshold of the size reduced ratio. A strip will be cropped if its reduced ratio exceeds the threshold to prevent yielding local warping and preserve salient content of the image. Experimental results show that the proposed algorithm can protect both the local structures and the global visual effect of the image, yielding satisfactory resizing results.

Note:新增回溯数据