The automation of AgriFood quality control processes requires computing systems running image processing algorithms, whose execution can be benefitted significantly by implementation in digital hardware (HW). In this context, image processing tasks have been successfully implemented using models of cellular automata (CA). Any image can be viewed as a two-dimensional (2D) CA, whose inherent massive parallelism facilitates the acceleration of the model's execution when implemented in HW. In this work we present results from the design and the implementation on FPGA of a digital system running CA-based algorithms for erosion and edge detection on binary images, targeting quality control processes in industrial production lines. The system's performance was evaluated on a dataset of images of fruits. We present the key considerations for the implementation of such a CA-based approach to image processing tasks in digital HW, compared to alternative software-based solutions. The experimental results prove the correct operation of the system and demonstrate adequate precision even in small image resolutions up to 100x100 pixels.