A genetic algorithm approach for image representation learning through color quantization

Over the last decades, hand-crafted feature extractors have been used to encode image visual properties into vectors. Recently, data-driven learning approaches successfully explored as alternatives for producing more representative features. In this work, we combine both research venues, focusing on color quantization problem. We propose two learn representations through search optimized schemes, which lead effective extraction algorithms and compact representations. Our strategy employs Genetic Algorithm, a soft-computing apparatus utilized in Information-retrieval-related optimization problems. hypothesize that changing affects quality of description approaches, leading efficient evaluate our content-based retrieval tasks, considering eight well-known datasets with different properties. Results indicate approach focused representation effectiveness outperformed baselines all tested scenarios. The other approach, also considers size created representations, produced competitive results keeping or even reducing dimensionality vectors up 25%.