The Smoke Detection Method Based on Structural Similarity of Video

摘要

The structural similarity (SSIM) of image is generally used for objective quality assessment of image. The background image blurred by smoke will cause the degradation of image quality. In this paper, we discuss qualitative relation between SSIM index of smoke video and smoke density, and then propose a smoke detection method based on structural similarity of video. The experimental results demonstrate that the increasing of smoke density results in the regularly decreasing of smoke video's SSIM index. The difference of SSIM index between smoke and non-smoke distractors, such as moving people, provides an effective way to distinguish them.