A method and system for combining a set of n still image frames into a single reduced noise still image. The invention reduces visible noise, along with some television color artifacts, in still images captured from a video stream by combining data from multiple video frames. Each frame comprises an XY array of picture element (pixel) data. In the preferred embodiment, each pixel includes data defining an intensity level as well as color values. A weighted average of several frames of video is taken to generate a single still output frame. The weighting is determined adaptively in order to reduce inter-frame motion artifacts. The resulting output has substantially less visible noise than a simple unweighted average. The algorithm is much faster and simpler to implement than true multi-pixel motion estimation, which is not really necessary in a series of nominally-still images.