A METHOD FOR SEPARATING THE IMAGES OF OUTSIDE SCENES USING PERCEPTUAL ORGANIZATION AND BACKGROUND RECOGNITION

Abstract

Using color and texture information, a new outdoor scene image segmentation method based on background recognition and perceptual organization is utilized to identify background items including the ground, sky, and plants. A perceptual organization model was created for structurally challenging objects, which typically have several constituent parts. This model is able to identify the non-accidental structural relationships between the constituent parts of the structured objects and, as a result, group them together appropriately without requiring prior knowledge of the particular objects. According to the experimental results, the suggested method obtained correct segmentation quality on a variety of outdoor natural scene contexts and outperformed two cutting-edge picture segmentation techniques on two difficult outdoor databases (the Berkeley segmentation data set and the Gould data set).