Visual representation of an environment in topological maps is a challenging task since different factors such as variable lighting conditions, viewpoints, mobility of robots, dynamic and featureless appearance, etc., can affect the representation. This paper presents a novel method for appearance-based visual topological mapping using low resolution omni-directional images. The proposed method employs a pixel-by-pixel comparison strategy. Successive images captured as a mobile robot traverses its environment are compared to estimate their dissimilarity from a reference image. Specific locations (nodes in the topological map) are then selected using a variable sampling rate based on changes in the appearance of the environment. Loop-closures are created using a dynamic threshold based on variability of the environment appearance. The method therefore proposes a full SLAM solution to create topological maps. The method was tested on multiple datasets, which were captured under different weather conditions along various trajectories. GPS coordinates were used to stamp each image as ground truth for evaluation and visualisation only. We also compared our method with state of the art feature-based methods.