We propose a new spectral index, the Normalized Difference Fraction Index (NDFI), for enhanced detection of forest canopy damage caused by selective logging activities and associated forest fires. The NDFI synthesizes information from several component fraction images derived from spectral mixture models. Interpretation of the NDFI data is facilitated by a contextual classification algorithm (CCA) that enables accurate mapping of logging and fire-derived canopy damages. The CCA utilizes detected log landings, which are the spatial signature of selective logging, as starting locations for searching the NDFI image for canopy damage. This process separates canopy changes due to logging and associated forest fires from those caused by other natural disturbances. These methods were tested in the Sinop region, in the Southern Brazilian Amazon. Forest transect inventories, conducted along a gradient of degraded forests, were used to evaluated the performance of the NDFI. The NDFI was more sensitive to canopy damage than any individual fraction and is shown to have the potential for further sub-classification of degradation levels in forest environments. Map accuracy of forest canopy damage using the CCA classifier, assessed with aerial videography images, was 94%. The proposed NDFI-CCA classifier approach can be fully automated and, therefore, holds great promise as a forest monitoring tool in tropical forests.