Study On The Characteristic Detection Algorithm Of Static Image Data

Vision is the most direct,important and most effective way for human beings to perceive the world.This way enables humans to perceive the world well.Our daily interactions with the external environment are always in place,and a large part of these interactions come from the visual perception of the human eye.Visual attention mechanism,the selective attention mechanism is unique to humans and other primates animals,which makes the human can use such as color,shape,location of multiple low-level features,to realize the visual perception.First,the image's significant center is positioned,and then several low-level features are obtained in parallel processing.And then through the brain's combination to realize the characteristics of the integration,there will be a lot of background noise and interference to eliminate and suppress,finally obtain significant areas of object images or is interested in area.With the rapid development of communication technology and Internet technology,especially the improvement of the computing power of hardware,the development of image processing and machine vision technology has been greatly promoted.Related to significant project detection model for computer rapid interested in selecting the image area in order to more effective image processing is put forward a claim for higher faster and more accurate,prompt significant areas in the field of computer vision detection become a research hotspot in^[1].A large number of scholars have put forward many effective detection theories and models through hard work.This article is standing on the shoulders of giants do further explore,through consulting a large number of literature,the existing classic algorithms thoughts have a profound understanding of,and in these algorithms,including qualitative and quantitative aspects of the in-depth analysis,found some shortcomings and defects.In this paper,a new approach is proposed to try to improve the visual significance detection algorithm.According to the Adaboost algorithm,the misclassified samples were strengthened and then retrained,and each training was added to the idea of a classifier.In this paper,the low-level features of images are used as different weak classifiers of Adaboost,and the iterative fusion is carried out.In an iterative process in order to determine the coefficient of each weak classifier,introduce the thinking of information entropy to quantify uncertainty random variables,to calculate the information entropy of all the weak classifier to modify the coefficient of each weak classifier.Distinguished from the classical significance test model,we mainly did the following work:?1?The details of the target are reinforced.A large number of existing algorithms are tested by using global static characteristics alone or using local static characteristics independently.This only gets a rough target mark.Therefore,this paper starts from the static feature extraction and adopts the method of iterative fusion introduced by the traditional feature fusion theory,and USES the parallel method to realize the significance area detection.First of all,the feature of image is used to include the local shape feature of the global color feature,and to some extent,it improves the information loss caused by the traditional single feature.Then,the acquired image feature is used as a weak classifier,and the Adaboost idea is used to enhance the weakly classifier into a higher precision classifier.The experimental results on the public image set show that the model presented in this paper is better than other classical models in the mean absolute error and other indexes.?2?The significant detection model based on multi-static feature parallel fusion has achieved good results in the detection of the significant area of the image.In the use of Adaboost algorithm,the automatic allocation parameter is learned to realize the weak classifier coefficient selection more effectively.Due to feature partition is not enough detailed in the integrity of the significant target detection and the influence of noise,using the information entropy and weighted average calculation,determine the influence of different static characteristics of the weight,will be the detail place of the significant areas to further strengthen.The center-edge feature is used for noise suppression.The experimental results on the public image set show that the method adopted in this paper is applied to the model,and the average absolute error and other indexes are better than other classical models.