Regularized restoration of partial-response distortions in sporadically degraded images

The complex image degradations observed in (video) image sequences are often due to partial-response mechanisms caused by the physical limitations of practical cameras. The restoration of these distortions require more sophisticated image models that consider the underlying image acquisition process. In this work, a broad class of spatially varying distortions is defined. The observed image is modeled as the physically meaningful superposition of K partially degraded images. In areas where the degradation is due to the partial-response of the sensor to multiple object (regions), the distortion is said to be clustered or sporadic. The sporadic distortions under consideration here are spatially varying, object dependent and difficult to estimate from the observed data. A regularized iterative restoration of sporadically degraded images resulting in K restored images is proposed. The algorithm demonstrates a promising (object based) image restoration approach for video editing applications