The assessment of fragmentation through blasting and therefore subsequent crushing and grinding stages is important in order to control and optimize the mining operation. Prediction of the mean size of fragmented rock by the rock mass characteristics, the blasting geometry, the technical parameters and the explosive properties is an important challenge for the blasting engineers. Some of the effective parameters on rock fragmentation have been investigated in several empirical models. A model for fragmentation in bench blasting was developed using the effective parameters on the existing empirical models, so as to propose a simple applicable model to predict X50. The proposed model was calibrated by nonlinear fits to 35 bench blasts in different sites from Sungun copper mine, Akdaglar quarry and Mrica quarry. In order to validate the proposed model, its results were compared to data obtained from six blast sites in Chadormalu iron ore mine and one in Porgera gold mine. The results indicated a small variance in X50 which is calculated by the proposed model, through the image processing approach. The Comparison of the powers between the proposed and the Kuz-Ram models showed that the specific explosive energy and the powder factor are almost the same. The advantage of the proposed model rather than the Kuz-Ram model is specific explosive energy, since this parameter includes the powder factor and the weight strength of explosive. Also a sensitivity analysis was carried out based on artificial neural network. The results showed that the burden and the specific explosive energy were the most effective parameters in the proposed model.