Research And Application Of Weak Monotonic Averaging Image Reduction Algorithm

With the continuous innovation of electronic information technology,high-definition image and video capture devices have been widely used in various fields of life,which provided a great deal of convenience for peoples.However,the images captured by these devices either cannot be displayed in full size on the small display devices,or cannot meet specific archiving and retention requirements then required the image reduction techniques to preprocess these images.Traditional monotonic averaging image reduction operators,such as the arithmetic mean reduction operator and the median reduction operator are susceptible to noise interference when the images are reduced,and cannot retain the details information of the images after reduction.Although the existing pixel cluster compactness based weak monotonic averaging reduction operator has better small feature retention ability,it needs to pre-specify the background color of the image to be reduced,if the pre-specified background color is inconsistency with the image to be reduced the weight calculation error will be generated,then lead to the decrease of the reduction performance.To this end,this thesis combining the pixel distribution statistics and mapping functions proposed an improved pixel cluster compactness based image reduction operators.This operator can achieve the background self-adaptive weight calculation by utilizing pixel distribution in the image to model the probability of the background color,which avoiding the problem of the performance degradation caused by manually background color setting.The experimental results show that compared with the traditional pixel cluster compactness based weak monotonic averaging image reduction operator,it has better robustness and suitable for the reduction task of two different color background images in black and white.In order to further verify and illustrate the availability and practicability of the proposed improved pixel cluster compactness based weakly monotonic averaging image reduction operator,this thesis applies it to the field of video foreground extraction and analysis,and proposes a video foreground extraction framework which combined the improved pixel cluster compactness based weakly monotonic averaging image reduction operator and Gaussian mixture model.The experimental results show that introduce the improved pixel cluster compactness based weakly monotonic averaging image reduction operator to the Gaussian mixture model based video foreground extraction and analysis process can increase the anti-noise performance of the original model,and can reduce the storage cost of the system without affecting the accuracy of vehicle statistics.