A machine vision system that is capable of sorting oysters by size and detecting irregular shapes was developed in our previous work (Lee et al., 2004). The vision system includes a custom-designed light chamber, special camera and optics arrangement, software for computer vision algorithms, specific control and delivery mechanism, and friendly graphical user interface. However, Learning Vector Quantization (LVQ) was used for shape evaluation and required the training of a Neural Network. A simple and accurate shape evaluation method is needed for commercial use.