Digital Image Correlation (DIC) is a full-field measurement technique that generally relies on brightness conservation principles. This work aims to analyze cases in which the speckle pattern vanished after heating, thereby severely altering the brightness. Sets of images acquired during \textcolor{red}{\bf 6} experiments were registered and then used to obtain optimal sets of reference images for each time step via Principal Component Analysis (PCA). The proposed methodology significantly reduced the gray level residuals and provided insight into brightness variations that occurred in the analyzed cases. It was possible to study cases with severe speckle pattern changes by updating the reference image when standard DIC did not converge and using the PCA methodology to obtain optimal reference states.