We present a spatially and temporally adaptive iterative image restoration algorithm. It extends an earlier technique using row action projection (RAP) by continuously varying spatial adaptation for applying updates and smoothing instead of the earlier segmentation based approach. The new algorithm is a subset of the method of projection onto convex sets and incorporates any a priori information about the blurred image or the blur in order to reduce ambiguity in the solution and accelerate the speed of convergence. Our algorithm generalizes the projection concept to column action projection (CAP). In RAP, an error in the estimate of a pixel of the blurred image is used to update all those pixels of the restored image that contribute to it. In CAP, a pixel of the restored image is updated using the errors in the estimates of all those pixels of the blurred image that receive contributions from that pixel. This algorithm adds a way of emphasizing high-SNR regions during restoration, reducing edge blurs by using a new edge-oriented smoothing, morphologically processing the local variance of the restored image, and temporally varying the restoration parameters to provide significantly better SNR improvements during restoration.