A novel Genetic Algorithms and SURF-Based approach for image retargeting

• We propose a Genetic Algorithm approach for image retargeting.
• We present new algorithm genetic operators for reproduction and mutation.
• We present a novel index for quality assessment.
• Our model uses feature extraction to guide its behavior for feature preservation.
• We apply vector quantization algorithms to improve our quality assessment proposal.

Seam Carving is a method which aims at retargeting images, i.e., adjusting input images into arbitrary dimensions by removing pixel paths (seams) with minimal visual impact. The standard method is based on an exhaustive searching method for minimal cost seams according to a pixel relevance function also named energy function. In this work, we propose a new method to retarget images based on both Genetic Algorithms (GAs) and a high-performance scale and rotation-invariant interest point descriptor, the well-known method, Speed Up Robust Features (SURF). Along with our proposal, we present a content-aware fitness function, a new individual modeling, as well as a new index for image retargeting quality assessment. According to the simulations carried out, our proposal is very handy for image retargeting and outperforms the Seam Carving in terms of the image quality assessment index in most results.