A Study Of Real-time Synthesising Of High Dnamic Range Images

High Dynamic Range(HDR) images can be realisticly representing brightness and color information of the scene in the real world. Because of the good characteristics of high dynamic range images, in recent years, which in industrial image processing, image-based modeling, machine vision, virtual reality, three-dimensional digital entertainment and other fields has been widely applied. In particular, HDR images has outstanding advantages in the global lighting, realistic rendering aspects. Currently, high dynamic image acquisition devices on the market is very expensive, and the low dynamic range image acquisition system is difficult to get all detail informations of the scene luminance, in view of this, we use modern signal processing techniques to merge multiple exposured low dynamic range images achieve a high dynamic range image containing luminance information in order to reduct realistic scenario mostly, it is a relatively high cost-effective high-dynamic image acquisition methods.There are mainly two ways of synthesising High dynamic range image. First using of images of different exposure times for solving nonlinear response curve of the camera, the curve is the camera’s mapping function’s inverse function which maps scene’s radiation values into image’s pixel values. Using this curve we can restore radiance of the scene, which is the pixel value of the high dynamic range image. The second is to select the optimal block from images of different exposure times to generate and fusion high dynamic range image, however, the image will have sharpening effects which need to be eliminated later,which is a complex operation. Since the first algorithm is more conducive to hardware implementation, we only need to calculate to the modified nonlinear response curve in advance and save it on the chip, the we can use the curve to synthesis high dynamic range images.Based on the extensive research of existing high dynamic range imaging algorithm and comprehensive consideration of computational algorithms and synthesis effects, we propose an improved non-linear response curve method about image fusion. Our paper aiming the strictly limited of algorithm execution time and resource-consuming aspects in hardware system, first carring out optimization of algorithm in order to real-time integrate of multiple-exposured low dynamic images in a way through a lookup table in hardware-level, and verifing synthesis process in the software. Then, based on FPGA development platform, using Verilog HDL language programming to build high-dynamic image synthesis hardware circuit module for real-time synthesis of high dynamic image, the final hardware implementation results confirm the effectiveness of our methods. The proposed multi-exposure HDR image synthesis method will reduce the threshold for high dynamic range imaging technology and has positive significance in popular it.