Existing Camouflaged object detection (COD) methods mainly focus on extracting target features or constructing boundaries for camouflaged objects. However, background features can be considered a form of deceptive information for both humans and computer systems. Effectively handling background information can greatly improve the saliency representation of camouflaged objects. Hence, we propose a novel emphasizing object-background difference network (EOBD-Net). Specifically, we first devise a multi-scale feature fusion module (MSFM) to obtain features at multiple scales, contributing to a comprehensive understanding of the target and its surrounding environment. Secondly, we propose an object-background difference module (OBDM). OBDM incorporates boundary features to enhance the delineation of object and background boundaries, facilitating better separation of target and background regions. Subsequently, target and background regions are spatially fused within the local receptive fields at each layer and channel-wise through dynamically adjusting the weights of each channel, constructing information-rich target and background features. Then, OBDM adjusts the semantic correlation between target and background by multiplying learnable scale parameters with target and background features, effectively distinguishing camouflaged information from the background. We design a multi-level feature fusion module (MLFM) to fuse geometric details and semantic information of different feature levels. A large number of experiments show that our EOBD-Net has obvious advantages compared with the most advanced COD methods.