Because of its affordability and high-resolution perception, visual odometry (VO) is frequently used for mobile robot localization. Its accuracy is, however, extremely susceptible to visual deterioration brought on by things like low texture, motion-induced blur, and poor illumination. A simple and understandable approach to VO confidence estimation in real-time using internal system metrics is presented in this paper. The suggested method uses VO failure indicators, temporal consistency of feature matches, and inlier ratios to calculate a normalized confidence score using a logistic function. In contrast to previous approaches, our method relies only on static visual input and does not require robot motion. The confidence score accurately represents VO performance, according to experiments carried out in four controlled indoor environments: highly textured, low-textured, motion-blur, and poor illumination. High scores (0.76 ± 0.05) are obtained in Highly textured environments, whereas significant degradation (0.06 ± 0.004) occurs in Poor illumination conditions. To enable adaptive localization based on visual quality, the suggested score can be incorporated as a dynamic weighting factor in sensor fusion frameworks like Extended Kalman Filters. The approach is appropriate for confidence-aware navigation in autonomous mobile robots, computationally efficient, and compatible with ROS 2.