Depth cameras have demonstrated their ability to provide accurate distance measurements of the surrounding environment in multiple application areas. Similarly, UAVs are increasingly being deployed in various real-world scenarios. However, these platforms often operate under constraints such as limited computational power and strict weight restrictions. Therefore, some UAVs use a depth camera to have depth data. As raw depth data often contain significant noise or inaccurate measurements at different scales as distance changes, image correction techniques are essential for unplanned obstacle avoidance. We examine multiple steps to enhance image data for UAV obstacle avoidance, including the calibration process, the selection of a hole-filling filtering method to improve depth image quality by reducing the number of invalid pixels. We also propose an image processing algorithm based on exponential decay that mitigates the root mean square error of the distance measurements within 3,000 mm, 5,000 mm, and 7,000 mm by 0.64%, 4.42%, and 6.53% respectively relative to ground truth measurements. Finally, the algorithm was deployed on an NVIDIA Jetson Nano board, achieving 27 FPS on a 30 FPS configuration. The results obtained in this study enhance the quality of image data from this type of sensor, contributing to improved performance in UAV image analysis in challenging environments.