Scene Segmentation Methods (P-0104) (MFN 3068)

Abstract

Scene segmentation and interpretation is an important area of computer vision. Image segmentation helps to visualize different segments of an image separately and extract the desired information from them. Segmentation is often described by analogy to visual processes as a foreground/background separation, implying that the selection procedure concentrates on a single kind of feature and discards the rest. Different techniques have been developed to segment the image like region based methods and clustering based methods, but we will use sequential region growing technique as required by the project. The main goal of using this sequential method is its simplicity and effectiveness in segmenting the image. This approach is fast and gives satisfactory segmentation results. Our proposed sequential region growing algorithm consists of following steps: 1) Preprocessing 2) Region Growing 3) Post Processing. Each step has several sub steps to achieve the desired goal.