Effective Moving Object Detection Using Background Subtraction in Stationary Wavelet Domain

Related Works

Background subtraction approaches generally work to find the absolute difference between the current frame and the background so that relevant changes can be detected. The algorithm's success is dependent on developing an efficient method for modeling and updating the background. Background subtraction works well for extracting all types of objects from videos, but the effectiveness of the approach deteriorates if the background is not static, has lighting fluctuations, and the videos are noisy.

A number of algorithms have been suggested for moving object detection from the standard background subtraction (BS) method and the wavelet transform, which splits frame sequences into detailed and approximate components and performs other operations on only those components. Kavitha et al. (2017) proposed a new method using stationary wavelet transform (SWT) to identify and remove moving shadows based on a threshold defined by wavelet coefficients. The multi-resolution feature of the stationary wavelet transform decomposes the frames into four different bands without losing information. Cheng et al. (2006) based their approach on Discrete Wave Transformation Technology (DWT) as a preprocessing process for detecting and tracking moving objects. 2-D DWT was used to analyze the image into four sub-images (LL, LH, HL and HH). The LL3 range(band) is used for further detection due to Consider low computational costs and noise reduction issues. However, using haar-based DWT distorts the shape of the object. In this study (Chih H. et al, 2014), an improved BS approach based on the Gaussian mixture model (GMM) and wavelet transform (WT) is suggested for the challenge in object detection. Not only can the effect of lighting changes, noise, and shadows be reduced, but dynamic changes to landscapes can also be handled using this method. In Alok et al. (2014), a video surveillance method using the complex wavelet transform and the method based on the approximate median filter for the segmentation of moving objects.

According to the problem of detecting and tracking the moving objects of a video surveillance, the most used techniques deal with a stationary camera (Ismail et al 1998) or closed world representations (Isard et al, 1998) which rely on a fixed background or a specific knowledge on the type of actions taking place, where various difficult cases are not perfectly solved and must be improved such as occlusion, tracking, identification of object, localization and removing shadows of objects.