Concurrent Object Segmentation and Recognition with Graph Partitioning

Abstract

Segmentation and recognition have long been treated as two separate processes. We propose a mechanism based on spectral graph partitioning that readily combine the two processes into one. A part-based recognition system detects object patches, supplies their partial segmentations and knowledge about the spatial configurations of the object. The goal of patch grouping is to find a set of patches that conform best to the object configuration. This process is integrated with the pixel grouping based on low-level feature similarity, through pixel-patch interactions and patch competition that is encoded as constraints in the solution space. The globally optimal partition is obtained by solving a constrained eigenvalue problem. We demonstrate that the resulting object segmentation eliminates local false positives at the high level of part detection, while overcoming occlusion and weak contours at the low level of edge detection.