In this paper we use an approach to image enhance- ment using anisotropic diffusion filter applied in low-contrast de- fects in no homogeneous lighting situations, such defects are dif- ficult to detect with standard algorithms for segmentation due to their nature. The applied filter is based on extended anisotropic diffusion, with an enhanced operator which increases the con- trast of the defect; but with the possibility of doing softer in the background. The result of applying the filter shows a quantita- tive improvement in dark contrast for low contrast defects, which produces better results in later stages such as segmentation. We analyzed a synthetic image and several real cases of surfaces with dark defects. In both situations the proposed strategy worked. Human experts were used to qualitatively assess the effect of im- proving the segmentation results, we analyzed 140 real images, and it was found that the methodology effectively enhances low contrast dark defects; however, it should explore strategies to pre- vent enhancement of noise in the images.