This chapter examines multiframe image super-resolution in a probabilistic framework. Many multiframe super-resolution algorithms begin by a point estimate of the unknown latent parameters, such as those describing the motion or the blur function. The focus of this chapter is on alternatives to this practice that can yield superior super-resolution results.
Subhasis ChaudhuriFitzpatrick, BrianKang, HyeonbaeRuiz, MatiasYu, SanghyeonZhang, Hai
Zhengyu WuZhengjun LiuYutong Li