Data-Driven Parameter Choice for Illumination Artifact Correction of Digital Images

We propose a new procedure for image illumination correction with data-driven parameter choice. This aims at estimating the reflectance from corrupted version in which corruption is due to pointwise multiplicative artifact. The log -illumination artefact consists of “smooth” variations intensity are modelled by function lying finite dimensional space. Then -y-correction incorporated. question model selection difficult solve. an entropy minimization criterion both approximating space dimension and -y-coefficient, so that no tuning needed. Several experiments presented using this approach. A comparison other methods illustrates relevance