Dual Transition Region Extraction based Colour Image Segmentation: Application to Fish Image Segmentation
Image segmentation using transition region has been quiet effective in recent years due to its simplicity. Previous approaches using transition region only concentrate in segmentation of gray scale images. Colour image segmentation using transition region approach is a challenging task due to the increase in complexity involving various colour components. Here we have proposed a hybrid transition region approach for colour image segmentation. Two existing transition region based approaches: (i) Gabor based transition region approach and (ii) local variance based transition region approach are used to develop the proposed method. Initially, the R, G, B colour components are separated from the original image. Gabor based transition region approach is applied to segment the texture features from the image. The result of previous method is used as input to local variance based transition region approach for final object extraction from image. The proposed method works effectively on variety of images containing both single and multiple objects. The method is applied for fish image segmentation. Experimental results revel that the proposed method outperforms many existing approaches.
