Hybrid bi-dimensional empirical mode decomposition based enhancement technique for extreme low contrast UAV thermal images

Abstract

The performance of automatic target detection and classification systems are typically affected by reduced contrast quality introduced by external interferences. In particular for Unmanned Aerial Vehicle (UAV) captured thermal surveillance images, the effect is more evident. This advises the use of contrast enhancement technique as a solution to enhance the reduced contrast of hot regions for efficient target detection. In this paper, a simple and novel enhancement technique based on singular value decomposition (SVD) using Bi-Dimensional Empirical Mode Decomposition (BEMD) is proposed to enhance the hot regions in extreme low contrast thermal images captured by UAV. In the first step, the technique decomposes the thermal image into Intrinsic Mode Functions (IMFs) and residue by using BEMD. In the second step, it applies Contrast Limited Adaptive Histogram Equalization (CLAHE) in the residue for local contrast enhancement and then calculates the singular value matrix. In the third step, residue component is rescaled for further improvement of hot regions using scaling factor. In the fourth step, a detail enhanced IMF components are generated using gray scale transformation. Finally, the contrast enhanced residue undergoes Inverse BEMD (IBEMD) together with the detailed enhanced IMFs for enhanced image generation. Experimental results demonstrate that the proposed technique effectively enhances the contrast and details in the image with less visual artefacts than other state-of-the-art techniques.