A method for semi-automatic chromatic equalization on multiple images in GRASS

This paper presents a method for color equalization of a n-number of images using a semi- automatic regression-based method. The images equalization will be based on contiguous images where one will be considered the “master” and the other the “slave” depending on which will be used as reference and which will be changed respectively. The idea behind this procedure is to match user-input-points which are supposed to have the same chromatic characteristics. The problem of unsightly patch lines often arises when texturing 3D models with digital images taken with different lighting conditions and camera settings. The example which is used shows the validity of such procedure. It is a Chapel by Giotto in Padova, Italy called “Cappella degli Scrovegni”. After doing a 3D model from a laser scan inside the structure, digital images of the frescoed chapel walls where draped over it thus applying texture. The textured model presented odd color differences especially visible along the patch-line between two images which were draped next to each other in the model. Our method implies for the operator to add control points which the program identifies as having to be have equal chromatic characteristics. It then uses grey levels for the three RGB color bands around the points to bring frequencies distribution of the slave to the master and thus compute a trasformation look up table for each pair of points. Grass's characteristics of scripting and open-libraries permit the implementation of such a module. This open source program will therefore be used as the base tool for programming