Most video exploitation algorithms operate on individual frames. To effect good results in such applications, the algorithms require good frames with which to work. However, videos may contain artifacts such as blur which inhibit the extraction of inherent and useful information. This paper proposes an algorithm that detects poor video frames induced by global motion blur. The proposed algorithm is divided into two steps: the first of which creates a single image blur metric, and the second of which adds temporal information. The blur metric is derived from a linear least squares fit to the log distribution of the highest subbands in a wavelet-based Haar filters. The second part of the algorithm correlates adjacent frames to boost performance. The ideas presented in this paper are low in complexity yet high in performance. Additionally, the proposed algorithm has been tested on natural video data as well as synthesized blur, and comparisons to state of the art show an advantage in using wavelet-based thresholding.