Realization Of Video Image Sharpness Detection Module

With the promotion of the city's peaceful urban construction,the number of medium-sized urban surveillance system cameras is as many as tens of thousands,how to assess the quality of these surveillance systems has become an urgent problem to be solved.Video image sharpness detection performance is an important basis for the quality evaluation of urban video surveillance system and one of the current research hotspots.This thesis first introduces a framework of quality evaluation system for urban surveillance system.Among them,video image sharpness detection module is one of the important bases of the system framework.Secondly,the traditional video image sharpness detection is mainly based on the frequency domain or spatial domain blur feature of the video image.It can not meet the requirement of the actual system for the video adaptability of the actual urban area surveillance system.A blur index with blur feature in frequency domain and spatial domain is proposed.Then,for the problem that structural similarity index can not be well consistent with subjective blur evaluation results(SSIM blur index does not have a good monotonicity),the thesis improves the performance of the SSIM index appropriately by adding a saliency area detection algorithm based on the original algorithm.Finally,based on the detection of blur of single-frame video image,the realization scheme of video sharpness detection module of real video surveillance system is given.For different situations with still object or motive object in video scene,different video image sharpness algorithm is applied for detecting video sharpness.The accuracy of the video sharpness detection is further improved by adding the person and vehicle classification algorithm and identification algorithm.The experimental results show that the performance of the video image sharpness detection module has reached the expected design requirements.