Block objects detection based on entropy of brightness

For image compression in automatic target recognition(ATR),the design of detecting region of interest(ROI) is one of the key issues to be solved.The entropy of brightness is small when the change of brightness is severe and is large when the change of brightness is smooth.And it is relative to the mean of brightness.A novel method of detecting block objects based on the entropy of brightness is presented.The method can detect tanks,plans or ships in visible or infrared images effectively.Extensive experiments make it clear that the method is robust and adaptable,excelling the ROI detector based on watershed and satisfying the demand of ATR-directed image compression.