Abstarct: Human behavior analysis and visual anomaly detection are important applications of video surveillance systems, in fields such as security systems, intelligent houses, and elderly care. People re-identification is one of the main steps in a surveillance system that directly affects system performance. This step is used to determine labels of people in images considering their visual appearance. It is a challenging task as the appearance may change across the camera’s network. Body occlusion, crowded background, and variations in scene illumination and pose are some challenging issues in people re-identification. Previous re-identification approaches faced limitations while considering appearance changes in their tracking task. In this research, considering body occlusion caused by carried objects and crowded background, four weighing mechanisms are proposed to improve the performance of re-identification approaches. In each proposed weighing mechanism, a weight map is proposed to tune the effect of background and foreground regions on the extracted appearance characteristics proportional to the region’s importance in re-identification. In the first approach, using the Gaussian distribution and the probability density function, an unsupervised semantic segmentation approach is proposed, where it assigns pixels with the same importance to the same segment. The segmented image is then used as a weight map to extract appearance features form the image’s pixels in the weighted form. Due to some errors in the first approach, in the second weighing mechanism, using DeepLabv3+, the image is semantically segmented into three regions as: person body, possible carried objects, and background. Then, a constant number is assigned to the pixels associated with each region depending to its importance in people re-identification. To avoid weighing regions manually, in the third mechanism, a weight map is computed representing the importance of each region in the re-identification, depending to its size and location. Besides, considering the importance of retrieving the occluded parts of the body in re-identification, in the fourth approach, a pre-processing step namely unification process is proposed to retrieve the occluded parts of the body using their neighboring pixels. Then, a weight map is introduced in order to tune the effect of each pixel of the processed image on the extracted features considering its distance from the center of the image. Our proposed weighing mechanisms along with robust descxriptors, such as the Gaussian of Gaussian (GoG) and Hierarchical Gaussian Descxriptors (HGD), can enhance existing methods in dealing with the challenging issues in re-identification. Experimental results on VIPeR, PRID450s, CUHK01, and CUHK03 datasets demonstrate effectiveness of the proposed techniques in improving performance of existing re-identification methods.