The current work proposes a methodology for the recognition of plant species using features obtained from digital leaf images which are fragmented. Due to various environmental and biological factors leaves are fragile and prone to fragmentation. The paper studies how recognition of leaves can be effectively done when only a portion of the leaf can be obtained. The situation is considered: major fragmentation, where less than 25% of the leaf area is present. Since part of the leaf contour would be missing, shape based techniques will not be effective for recognition, due to which a combination of color and texture features using Normalized Cross Correlation have been used. Experimentations involving classification of 1500 fragmented leaf slices varying in shape, size and fragmentation, to 30 pre-defined classes, demonstrate effectiveness of the approach. © 2017 Taylor & Francis Group.