Study On Portrait Photography Supporting System Based On Lighting Perception

Photography is the art of lighting.How to control the light does affect the quality of photos.In order to create impressive masterpiece,experienced photographers adjust the exposure parameters by the Exposure Triangle to control the amount of light and the exposure time.Controllable operations are provided to experienced users and abundant photographic modes are provided to non-experienced users by modern complex digital cameras.With the development of electronic techniques and the population of portable digital devices,more and more photographic scenes in which ordinary users are able to take portraits by mobile devices pop up.However,cameras embedded into the mobile devices are not equipped with the ability to analyze photographic scene and the adjustable exposure parameters to control the light,resulting the poor photo quality.Based on the extensive research about the related work,we propose in this thesis a novel portrait photography supporting system,aiming to solve the problems occurred in the practice of portrait photography.The core of our system lies in the scene analysis algorithm and lighting perception algorithms both in the front-light mode and night mode.Our thesis firstly researched on the existing scene analysis algorithms.The scene analysis is referred to as the analysis of the lighting conditions in current circumstance during the photographic process,in order to locate the light source and measure the ambient lighting intensity quantitatively.Based on this study,we proposed the concepts that typical scene modes can be classified into front-light mode and night mode and the method to do that classification.Next we proposed the lighting perception algorithms both used in the front-light mode and night mode.This algorithm is named as the lighting perception because it is established on the analysis of lighting and ambient light.A lighting source tracing based on machine learning and an ambient lighting intensity measurement algorithm are proposed.Taking advantage of the lighting characteristics projected onto the subject's face by lighting source,a relationship between the lighting location and lighting characteristics has been established in order to trace the lighting source.In lighting perception algorithm,the model is established upon the circumstance that lighting projecting light onto the subject's face,on which creating unique lighting characteristics.A machine learning algorithm is exploited in this thesis to categorize the illumination maps according to the training dataset.A second lighting perception algorithm is also proposed in order to solve the problems occurred in the portrait photography in night scenes.The algorithm is able to determine the relationship between the ambient lighting and shooting distance and provide users with appropriate shooting distance when taking photos in dark scenes.The systematic implementations of the above-mentioned two algorithms,as well as the detailed description of interface with users when embedded into the mobile devices,are explained in the following part of the thesis.Finally,abundant experiments are carried out to evaluate the validation of the proposed system in different circumstances.