In this paper, we generate two novel colour component features (ClrComFea) for brain tumor segmentation from MRI images. This feature extraction process is purely based on the knowledge of multimedia colour models and MRI brain tumor characteristics. Initially the original grayscale MRI brain images are converted into colour images using pseudo colouring process. Then the pseudo coloured brain images are analysed for tumor features based on the presence of colours within the tumor locations. This subject specific knowledge about the brain tumor and colour components is used to generate two ClrComFea. Finally these two extracted features are segmented and treated as brain tumor and necrotic regions respectively. Experimental study was done over a set of T2-weighted MRI images. The validation process by the proposed method results in good quality and quantitative measures.
