Multiscale morphology based illumination normalization with enhanced local textures for face recognition

Morphology-Based Banknote Fitness Determination

“…The CDI method was developed to extract largescale soiling patterns, which required fixed size reference images. Also, the GDMDI, a modified version of the GDMQI method [13], was proposed to capture the small-scale soiling patterns. By combining these two methods, we developed a method that showed promising results for banknote databases that contained 18 different designs of EUR and RUB banknotes.…”

“…To solve this problem, the closing operation needs to be selectively applied to preserve strong edges so that different images have soiling patterns only. One possible solution is to use the generalized dynamic morphological quotient image (GDMQI) [13], a multiscale morphology-based method, which was first introduced for face illumination compensation. For face images, bright light sources may produce strong edges (shadows) on face surfaces.…”

“…One of the main concerns for most illumination compensation methods is to eliminate the effect of shadows. In this context, the GDMQI method can take advantage of the edge-preserving property of multiscale morphological operations [13], [16], [17], whereas other methods suffer from halo effects that usually occur around the shadow regions. The GDMQI is defined as follows:…”

“…Those are usually in small-scale and appear as dark patterns. We used a generalized dynamic morphological difference image (GDMDI) to extract such patterns, which is a modified version of GDMQI [13]. In this method, alignment errors can be ignored since strong edges can be suppressed with GDMDI.…”