主 講 人：陳定立 助研究員 (中央研究院統計研究所)
講 題：Image Denoising by Markov Random Field
日 期：98年12月01日（星期二）下午2:30 –3:20
地 點：數學系(科學館S433室)
摘 要：
We build a Markov Random Field (MRF) for image priors. The energy function of
the MRF is modeled as the sum of filter responses in each location. The filters
are trained by Principal Component Analysis (PCA) from an image database. The
coefficients of the MRF are estimated by maximum pseudo-likelihood method. Based on
this model for image prior, we take a Bayesian approach to recover noisy or damaged
images. The posterior distribution is calculated according to the image prior and the
observed image, and the desired solution is the image that maximizes the posterior
distribution. This solution can be obtained numerically through gradient ascent on the
logarithm of the posterior. In the end, we will compare our results with other existing methods.