HaraliCU: GPU-Powered Haralick Feature Extraction on Medical Images Exploiting the Full Dynamics of Gray-Scale Levels

摘要: Image texture extraction and analysis are fundamental steps in Computer Vision. In particular, considering the biomedical field, quantitative imaging methods increasingly gaining importance since they convey scientifically clinically relevant information for prediction, prognosis, treatment response assessment. this context, radiomic approaches fostering large-scale studies that can have a significant impact clinical practice. work, we focus on Haralick features, most common descriptors. These features based Gray-Level Co-occurrence Matrix (GLCM), whose computation is considerably intensive images characterized by high bit-depth (e.g., 16 bits), as case of medical detailed visual information. We propose here HaraliCU, an efficient strategy GLCM exhaustive set features. HaraliCU was conceived to exploit parallel capabilities modern Graphics Processing Units (GPUs), allowing us achieve up \(\sim \!20\times \) speed-up with respect corresponding C++ coded sequential version. Our GPU-powered solution highlights promising GPUs research.