Seam Carving Algorithm Based on Partitioning

The retargeting technique is essential for image resizing and greatly preserves regions of interest in the image. In view of the long operation time and the distortion of salient objects of the existing retargeting algorithm when many seams are removed, a new seam carving (SC) algorithm based on image blocking is proposed. Firstly, combine the Sobel algorithm with label decoupling framework (LDF) to fully extract the regions of interest in the image. Then image blocking reduces the range of dynamic programming, which can effectively reduce the running time of image retargeting. Finally, remove the seams in proportion to obtain the result. Experiments are conducted with RetargetMe and DUTOMRON image datasets. Experimental results show that, the proposed method preserves the integrity of significant regions, reduces the running time and improves the operational efficiency.