In recent years the quoit filter processing, that is a kind of filtering processing of mathematical morphology, has been developed as a technique to automatically extract the candidates of lung cancer lesion by a lung cancer examination system using X-rays CT, but there are some problems such as low process speed and the extraction accuracy. In this study we proposed some new image recognition methods for the purpose of raising identification rate by the image processing technique as a remedy of these problems. Firstly we analyze the characteristics of sample images of the lung cancer and confirmed the effectiveness of the methods we propose.