融合小波变换及SIFT算法的去抖动运动目标检测.

Unstable video images caused by camera jittering will lead to inaccurate or error detection results. Aiming at this problem, a motion compensation scheme is proposed by combining wavelet transform and adaptive SIFT algoritlim. First, the wavelet transform based on threshold is applied to remove noise of the image, the iterative criterion of variable step is established to adaptively search the matched SIFT key points, the match of features obtained is much more precisely. Then the jitter parameter is estimated by analyzing offset of these matched points. Finally, the Gaussian mixture models combined with wavelet transform are employed to detect moving target. Compared with the similar algorithm, the proposed method can obtain more precise and stable matched feature points such that the effective estimation of jitter parameters is assured. It can solve the jitter effectively and detect the moving target precisely and efficiently. [ABSTRACT FROM AUTHOR]