MODELING THE USER SATISFACTION AND PREFERENCES FUNCTION IN IMAGE ANALYSIS BASED ON THE LINEAR PROGRAMMING

The paper presents the method of modeling the user satisfaction and preferences function in image analysis. Function is modeled with use of linear programming. The aim of the function is to score the user pictures basing on the image attributes. Image attributes are obtained from image processing algorithm and images with high attributes values are tagged using 14 specified tags. To obtain the user satisfaction function weights a questionary method is presented, where k specially chosen photos are presented and user can chose any number that he/she finds attractive. The images are either obtained from user’s album or are generated in such way that maximizes the diversity of tags. The images are chosen with use of heuristic function that is presented and described. The results are discussed. The presented approach provides the compliance of user satisfaction model with reference on the level up to 83%. It has been shown that the more images are selected, the satisfaction function becomes more accurate, but the compliance value tends to a limit. This research is used for automated generation of travel photo album in such way that it personalizes the selected images according to individual preferences.