Design And Implementation Of Fast And Robust Muti-Exposure Fusion System For Mobile Phone

The quality of mobile phone image shooting is an important indicator for measuring the performance of the mobile phone.At present,the smartphone is limited in image shooting:On the one hand,due to the influence of the hardware sensor of the mobile phone,it is impossible to show the natural scenes observed by the eye;On the other hand,the images taken under complex shooting conditions have many problems such as overexposure and ghosting artifacts,which seriously affects the quality of the image.Multi-exposure image fusion technology generates high-quality images with bright colors and realistic visual effects by fusing different exposure images,which proves its effectiveness of improving image quality.According to the characteristics of mobile phone image shooting and current research status of multi-exposure image fusion,this thesis designs a multi-exposure image fusion algorithm for mobile device and implement a mobile phone multi-exposure fusion system based on this algorithm to improve the quality of mobile phone image capture.Firstly,this thesis builds a new multi-exposure image dataset to evaluate the performance of the algorithm in the scene of mobile shooting.Secondly,this thesis proposes a multi-exposure image fusion algorithm combining traditional methods and lightweight convolution network.Specifically,this thesis compensates the image dynamic area through the detection of the image structure consistency to reduce dynamic artifacts;This thesis uses the traditional method to realize the preliminary integration of multi-exposure image sequences,and enriches the fusion image details through multi-scale feature extraction methods.Through using lightweight convolution and attention mechanism,this thesis applies different weights to the multi-exposure image sequences which is designed to further process the fused image and enhances the adaptive ability of the algorithm.By comparison with lots of multi-exposure image fusion methods,this thesis proves the algorithm proposed in this article has better performance from the aspects of subjective image visual effects and objective image evaluation indicators.Finally,based on the proposed algorithm,a multi-exposure fusion system for mobile phones was realized to meet the actual application needs of the quality improvement of images.