Superpixel-based refinement of low-resolution foreground segmentation

摘要

-28 SUPERPIXEL-BASED REFINEMENT OF LOW-RESOLUTION FOREGROUND SEGMENTATION Disclosed is a computer-implemented method (300) of performing foreground 5 /background segmentation of an input image (370). The method receives a first foreground/background segmentation at a first resolution (360) of the input image and determines (320) a plurality of labelled (385) seed points (380) based on the first foreground/background segmentation of the input image. The method associates (330) each of the plurality of pixels in the input image with one of the determined labelled 10 seed points to obtain a second foreground/background segmentation of the input image, and performs (340) foreground/background separation on the input image at a second resolution by classifying each of the segments of the second segmentation as one of foreground and background based on the label of the associated seed point. -I Source image Modification process 360 High-res (HR) input image Low-res (LR) Start FG classification 2 of the source 320 image*I Seed Placement Seed locations / Seed Labels Superpiels Superpixel segments 390 385 39 Pixel Level Classification 340 High-res FG classification 395 ( End Fig. 3