Nonlinear Image Recovery with Half Quadratic Regularization英文文献.pdfVIP

~ 932 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 4, NO. 7, JULY 1995 Nonlinear Image Recovery with Half-Quadratic Regularization Donald Geman and Chengda Yang Abstract-One popular method for the recovery of an ideal imaging system and, with white Gaussian noise, the random intensity image from corrupted or indirect measurements is variables {Y,,t E D } are independent and normally dis- regularization: minimize an objective function that enforces a tributed, with constant variance and means E() = (Kx”)t. roughness penalty in addition to coherence with the data. Linear estimates are relatively easy to compute but generally introduce Another example occurs in computed tomography in which K systematic errors; for example, they are incapable of recovering represents an attenuated Radon transform: the data are photon discontinuitiesand other important image attributes.In contrast, counts over an array of detectors D and follow a Poisson nonlinear estimates are more accurate but are often far less distribution. accessible. This is particularly true when the objective function is In general, the information provided by the data alone (or nonconvex,and the distribution of each data component depends on many image components through a linear operator with even the data together with the forward model) is not sufficient broad support. Our approach is based on an auxiliary array and to determine xowith acceptable accuracy. For