LfePy: a Python package for local feature extraction with CPU and GPU compatibility

Abstract

Abstract Local features identify and describe distinct patterns or features in images at a localized level. However, extracting features from images is crucial for image analysis, as it enables models to acquire knowledge and identify patterns. Therefore, we introduce a novel Python package, LfePy (Local Feature Extractors for Python), that utilizes several local descriptors to extract features from grayscale images, ensuring compatibility with both Central Processing Units (CPUs) and Graphical Processing Units (GPUs). The package encompasses a range of techniques for addressing computer vision and image processing challenges. The LfePy package contains twenty-seven histogram-based descriptors and other essential image-processing methods. The package achieves a fast processing time for extracting features from images, as it includes a Graphical Processing Unit (GPU)-based version that outperforms related packages. This package is designed to advance the field of image analysis and related areas. It offers versatility, enhances performance, facilitates research, and supports a wide range of applications.