In this paper, we propose a novel image deblurring framework, which noticeably improves the effectiveness and efficiency of state-of-the-art approaches. In professional imaging with its typical shallow depth-of-field, it is challenging to estimate the exact focus distance during
recording, which often implies costly re-shooting. For the correction of blurred material in post-production, there exist a few deblurring methods, which are, however, challenged by working on real camera data due to noise and the general ill-posedness of the deblurring problem itself. Since
the effective out-of-focus operating range of deblurring methods is small and the blur characteristics are strongly depth-dependent, we introduce a framework where a depth map and measured lens characteristics ingest into a selection of state-of-the-art deblurring methods. Therefor, we introduce
a depthdependent parameter selection concerning blur kernels and smoothing weights first. Second, using these parameters the out-of-focus areas are selectively deblurred in order to overcome the emergence of strong artifacts. The foundation for the provided evaluation is formed by a dataset
with eight real images captured with a cinematic RGB plus depth camera containing multi-planar and in-scene depth-varying image content. Therein, we show visually and numerically that introducing our depth framework improves the deblurring performance and suppresses typical strong artifacts.