Hi Mthembu, 

This are very good references related to your question: see the attachment

1. Generating global Leaf Area Index from Landsat: Algorithm formulation

and demonstration

2. Step by step procedure 

http://cybele.bu.edu/download/manuscripts/sganguly03.pdf

http://www.nateko.lu.se/Courses/NGEN05/LE/LAI_APAR_exercise.pdf

That depends on your accuracy requirements. You will ideally have a linear statistical model that relates sugarcane LAI to a vegetation index. Extending this model to new data will assume that the Landsat pixel values you are working with are the same as that used to derive the vegetation index applied to the statistical model (i.e., top-of-canopy reflectance, radiance). If you have such an equation then you just multiply the "new" landsat vegetation index values by the model slope and then add the intercept. The result will give you sugarcane LAI.