Point source illumination simulation and adaptive pixel-level correction for tunnel scenarios

Tunnels are critical infrastructure in urban transportation and underground engineering, where high-quality images are essential for effective operation and maintenance (O and M). Point source illumination is widely used for tunnel imaging due to its practical deployment and low energy consumption. However, their spherical radiation characteristics often lead to severe non-uniform illumination in captured images, manifesting as over-bright centers and under-bright edges. To address this issue, this paper proposes an adaptive pixel-level illumination correction algorithm. The method integrates the entropy measure of enhancement (EME) with a Gaussian filter to restore multi-scale non-uniform illumination. Specifically, an EME-based block partitioning strategy is introduced, and the key parameter of the Gaussian filter, the cutoff frequency, is determined adaptively according to local brightness characteristics from pixel to block level. This approach effectively mitigates the bright-center and dark-edge problem. Additionally, a point source illumination simulation method is developed to generate realistic and varied non-uniform illumination patterns, enabling the creation of a dedicated tunnel image dataset from 20 sharp rock mass images for algorithm validation. Both qualitative and quantitative experiments demonstrate that the proposed correction algorithm outperforms several mainstream methods.