Content-Aware Image Retargeting with Reflection-Aware Energy Augmentation and Boundary-Preserving Seam Optimization

Abstract

Seam carving (SC) deforms reflective regions and object boundaries by eliminating lowcost seams. To address this limitations, this paper integrates proposed augmented map (PAM) with proposed image morphology-based retargeting algorithm (PIMRA). The PAM is the fusion of multiple operators such as the base energy map, reflection likelihood map, and saliency map. The PIMRA comprises dilation and erosion operations for energy augmentation along objects boundaries and reflective areas. The integration of PAM and PIMRA is assessed against existing state of arts such as scaling, cropping, SC, and Multi-Operator. The comparative analysis is performed based on parameters such as brightness, contrast, structure, distortion on edges and gradients, noise, and reflectance alignment.  The user study presents the efficacy of PIMRA in preserving structure, naturalness, reflection realism, and adaptability. Results confirm that the PIMRA integrated with PAM effectively preserves reflective regions and maintains object boundary integrity during image resizing.