Research And Implementation Of Key Algorithms For Automatic Mobile Phone Album Arrangement

Mobile phone has become one of the necessary terminal devices,and taking photos is one of its key functions.With the increase of the service time,more and more photos are saved in the mobile phone album,which makes it an urgent problem to automatically arrange the mobile phone album.In this thesis the key algorithms for automatic mobile phone album arrangement are researched and implemented.Our work is as follows:1)A low quality image detection algorithm is desigedLow quality image detection includes high/low illumination images detection and blurry images detection.This thesis uses the dark channel prior algorithm to detect the high/low illumination images,and the local gradient algorithm is used to realize blurry image detection.Experiment results demonstrate that the mean accuracy of the designed algorithm can reach 95.8%.2)An automatic clustering algorithm for mobile phone album is desigedIn this thesis,the local features of the image are extracted,and the similarity is calculated with the multi-level feature points matching technology,the target scene is used as a clue to realize the scene clustering in the album.Experiment results demonstrate that the clustering purity of our algorithm is as high as 94.66%.3)A content based on image retrieval algorithm is desigedThe semantic information is used in this thesis,and we combines the adaptive measurement matrix with different feature space to get the algorithm measurement.The experimental results show the AUC index value of this algorithm is 96.1%,which is improved by 4.7%comparing with the traditional markov algorithm.4)The algorithms are successfully transplanted to the Andriod system.In this thesis,the algorithms are transplanted to Android development board,and test the main fuctions including of low resolution image detection,high/low illumination image detection,blurry image detection and scene clustering etc.