Enhancing Monocular Depth Estimation via Image Pre-processing Techniques

Abstract

Robots and drones are getting popular in many applications nowadays. Autonomous operations of drones and robots are highly desirable to minimize human interventions and enhance operation efficiency. However, there are several challenges that need to be overcome before robots and drones can be automated with minimum hardware requirements. Currently, robotics industry employs costly sensors such as Lidar to estimate distance between a vehicle and objects. Recent advancement in Artificial Intelligence (AI) encouraged researcher to investigate techniques to estimate the distance between vehicle and objects using monocular camera and AI. However, distance (depth) estimation using monocular camera still suffers from low accuracy rate in depth estimation. This paper aims to improve the depth estimation values through applying several image pre-processing techniques such as Nonuniform Illumination Removal, Local Adaptive Thresholding, Histogram Equalization, Adaptive Histogram Equalization, White Balance, and Homo- morphic filtering techniques.