This research explores methods of improving the seam carving retargeting algorithm to reduce artifacting and increase art-directability for content-aware image resizing. Integrating machine learning models for object identification and image segmentation can result in improved results as compared to naive seam carving. Further inclusion of NLP models into this workflow can result in natural-language driven control over the retargeted image. This proposed pipeline is versatile, and can easily integrate improved models as they are developed.

This research was published in MDPI Electronics, and can be accessed here: https://doi.org/10.3390/electronics13224459

This research was also presented at ACM SIGGRAPH 2023. The abstract can be accessed and cited here: https://dl.acm.org/doi/10.1145/3588028.3603671

The full thesis can be accessed at the Drexel University Digital Library here: https://doi.org/10.17918/00001699