Integrating low-level and semantic features for object consistent segmentation

摘要

The aim of semantic segmentation is to assign each pixel a semantic label. Numerous methods for semantic segmentation have been proposed in recent years and most of them chose pixel or superpixel as the processing primitives. However, as the information contained in a pixel or a superpixel is not discriminative enough, the outputs of these algorithms are usually not object consistent. To tackle this problem, we introduce the concept of object-like regions as a new and higher level processing primitive. We first experimentally showed that using groundtruth segments as processing primitives can boost semantic segmentation accuracy, and then proposed a novel method to produce regions that resemble the groundtruth regions, which we named them as 'object-like regions'. We achieve this by integrating state of the art low-level segmentation algorithms with typical semantic segmentation algorithms through a novel semantic feature feedback mechanism. We present experimental results on the publicly available image understanding dataset MSRC21 and Stanford background dataset, showing that the new method can achieve relatively good semantic segmentation results with far fewer processing primitives.