MATLAB Code for Shadow Detection Using Illumination-Invariant Transformation

This article discusses a shadow detection method developed by Mr. J.W. Hofstee and Mr. E.J. Hanten, presented in their paper "Shadow Segmentation Based on Image Transformation for Illumination Changes" at the International Conference on Agricultural Engineering in Zurich. The core technique utilizes illumination-invariant transformation for shadow detection, which has been extensively researched in computer vision applications.

The method's key innovation lies in applying image transformations that maintain detection accuracy under varying lighting conditions. Implementation typically involves converting RGB images to illumination-invariant color spaces and applying threshold-based segmentation. While the method demonstrates excellent performance on certain images with clear shadow boundaries, its effectiveness decreases on images with complex textures or gradual shadow transitions. Potential improvements could include adaptive thresholding algorithms and machine learning-based refinement steps to enhance robustness across diverse scenarios.