High Dynamic Range Imaging Algorithm Based On 4Cell Image Sensor And Its Optimization On Mobile Devices

At present,mobile devices such as smartphones have become the most widely used equipment for taking photos in our daily life.When using mobile devices to take photos,people often encounter some high dynamic range scenes.In this condition,the images obtained often have partial underexposure or partial overexposure.Therefore,high dynamic range imaging has become one of the most important research field of taking photos on mobile devices.At the same time,the 4Cell image sensor has become more and more popular in current mobile devices due to its excellent photographic performance in low light.In this dissertation,a high dynamic range imaging algorithm based on the 4Cell image sensor is proposed.Then,the algorithm is optimized and applied to mobile devices.First,in order to solve the problem of acquiring high dynamic range images,This dissertation proposes a new method of multi-exposure image acquisition with locally changed exposure parameters on the image sensor by taking the advantage of the structural characteristics of 4Cell image sensor.This method makes it possible to obtain a multi-exposure image through a single shot,so that it can suppress the occurrence of ghosts in the synthesized high dynamic range image theoretically.Then,in the algorithm of synthesizing high dynamic range images from multiple exposure images,RAW image is used for synthesis and image noise is suppressed in the synthesis process,which further improves the dynamic range and quality of the synthesized image.Experiments show that the HIGRADE score and image information entropy of this algorithm are higher than other synthetic algorithms in multiple scenes.Second,in order to enable high dynamic range images to be displayed on mobile devices,a tone mapping algorithm based on guided filtering is proposed,which maps a high dynamic range image to a low dynamic range image.This algorithm is an improvement on the original algorithm of tone mapping that is based on image decomposition.By using the guided filtering to perform image decomposition in the logarithmic domain of the image,the halo phenomenon of the original algorithm is reduced,and the algorithm is more efficient.Experiments show that the subjective visual effect of the tone mapped image is better,and the TMQI objective evaluation is better than other algorithms involved in the comparison.Finally,this dissertation integrates the algorithm into the camera application of Android mobile devices and optimizes the algorithm.In this dissertation,a fast algorithm of guided filtering is used to improve the efficiency of the algorithm and the parallelism of the algorithm is realized by the multi-core multi-threading and NEON instruction with the help of ARM processor.Performance test results show that the average acceleration ratio of the algorithm is 5.91,and the total execution time of the algorithm is less than 350 ms,which meets the needs of the mobile devices.