数字图像处理课件lecture6.ppt

* Moment Invariant Seven invariant moments are used This set of moments has been shown to be invariant to translation, rotation, and scale change * Algorithm of Moment Invariant Step-1: Calculate the moment of order (p+q) mpq Step-2: Calculate central moments ?pq Step-3: Calculate the normalized central moments ?pq Step-4: Calculate the invariant moments ?i , i =1,2,…7 * Given a 2-D continuous function f(x,y), moment of order (p+q) is given by: P and q are positive integers 0, 1,2, ... Uniqueness theorem A. Papoulis, Probability, Random Variables, and Stochastic Processes, McGraw-Hill, 1965. Step-1: Calculate the moment of order (p+q) - mpq * Uniqueness theorem states: if f(x,y) is piecewise continuous and f(x,y) has nonzero values only in finite part of the x-y plane, then moments of all order exist and the moment sequence mpq is uniquely determined by f(x,y) and vice versa. * The central moment is expressed as Step-2: Calculate the central moment - ?pq * For a digital image, we replace the double integral by a double summation. it is given by: * The central moments of order three are: ?00, ?01, ?02 ,?03, ?10, ?20, ?30, ?11, ?12, ?21 They are computed below: * Similarly, ?01 =0 * Please refer to reference book for others R.C. Gonzalez and P. Wintz, Digital Image Processing, Addison-Wesley, 1987. In summary * The normalized central moments are defined as, for p+q = 2, 3, ….. where Step-3: Calculate the normalized central moment - ?pq * The seven invariant moments are Step-4: Calculate the invariant moments - ?i * Examples Original Pattern Scaled to half size Rotated by 2° Mirror-imaged Rotated by 45° * Moment Invariants for the images on previous page * Summary of Extracting rotation-invariant features Ring-projection: Any object Implementation is simple Cumulative angular function-Fourier descriptor (CAF-FD) : Pattern is approximated by a boundary Implementation is simple Moment Invariant: Any ob