High Dynamic Range Imaging Algorithms

Abstract

Real world around us may contain high dynamic range (HDR) scenes but general purpose imaging devices available in the market have low dynamic range (LDR) imaging capability. Dynamic range of any image refers to the ratio of highest possible pixel value to the lowest possible pixel value. This thesis aims to study HDR imaging pipeline which builds a tone mapped HDR image using its multiple differently exposed LDR images. The pipeline consists of two basic algorithms: dynamic range expansion using differently exposed images and dynamic range compression. Thesis discusses their theories, mathematical models and implementation flows. Moreover, thesis introduces our proposed algorithm which helps in selecting appropriate tone mapping operator (TMO) from a TMO dictionary for any particular scene to be captured. The proposed algorithm uses tone mapped image quality index (TMQI) algorithm to evaluate the performance of each tone mapping operator maintained in our TMO dictionary. The importance of our proposed algorithm has also been outlined by comparing performance of each TMO of dictionary and then discussing the process of selecting the final output. The evaluation capability of TMQI algorithm has been demonstrated by surveying eight tone mapping operators’ performances over eighteen datasets.