Statistical indicator functions are developed that evaluate 10-cal hypothesis tests to detect the location of generalized boundaries within images. Functions based on T test, F test, and nonparametric Kolmogorov-Smirnov tests are developed, and implementation strategies for use as image processing operators are presented. The statistical indicator functions are compared to previously proposed indicator functions and the performance of each is evaluated in tests involving images corrupted with additive, multiplicative, and textured noise, as well as on real images. The statistical indicator functions are evaluated in a probabilistic framework, are sensitive to local context and, thus, detect generalized image boundaries with very low sensitivity to threshold selection. The test results show that the statistical indicator functions provide a more consistent response over a variety of image types than can be obtained using nonstatistical indicator functions. Statistical indicator functions are suitable for implementation in two-step image processing strategies such as those used for adaptive enhancement or adaptive quantization.