Particle size is considered one of the significant characteristics used in geotechnical practices. Traditionally, sieve analysis is utilized for coarse-grained soil. However, this method could be time-consuming and take much effort, especially for large scale infrastructure projects. This paper presents an efficient method for estimating gravel particle characterization utilizing image processing and artificial neural network technique (IPNN). The proposed algorithm is performed by utilizing particle boundary delineation and shape feature extraction to train a neural network model for estimating gravel size distribution curve. It is found that excellent agreement exists between the results obtained from conventional sieve analysis and neural analysis for gravel soil particles with maximum difference in passing percentages up to only 3.70%. The proposed technique shows satisfactory results for crushed stone samples with a maximum difference in passing percentages about 10.90% mainly in large-diameter particles. The presented technique (IPNN) could offer a promising alternative technique for material quality control process especially in large scale projects.