Image Semantic Segmentation Based on Multi-Scale Feature Extraction and Fully Connected Conditional Random Fields

Aiming at the problems of insufficient usage of context information and unclear image edge segmentation in image semantic segmentation, a network model based on multi-scale feature extraction and fully connected conditional random fields is proposed. RGH and depth images arc input into the network in a multi-scale form, and their features arc extracted by a Convolutional neural network. Depth information is added to supplement the RGH feature map and obtain a rough semantic segmentation, which is optimized by the fully connected conditional random fields. Finally, fine semantic segmentation results arc obtained. This proposed method improves the precision of semantic segmentation and optimizes the image edge segmentation, which has a practical application.