This paper studied the gray level dependence （GLD） texture analysis approach in both theory and application. GLD feature extraction was viewed as filtering of weight matrices on GLD matrices. The relation between image teyture and GLD matrices

and the influence of the shape of weight matrices on GLD features were discussed. Understanding of sensitivity and stability of GLD features has been improved. Selection of GLD parameters were studied in detail. Eight GLD features were evaluated in their performance and applications.