Image bit depth expansion

Abstract

Due to the increase demand of high dynamic range (HDR) content and increased difficulty to capture HDR videos, bit deth expansion techniques are used which maps the low dynamicrange (LDR) content to HDR providing high luminance for the viewers. To capture HDRcontent is not direct and as easy as capturing LDR content. It requires specialised equipment and thus leads to research for using the LDR content obtained by the existing cameras,Utilizing the existing content on HDR displays. These HDR displays have more number of bits per pixels then the standard LDR images, so it is necessary to increase the bit depth of the LDR content.In this thesis, a time efficient bit depth expansion technique is proposed for video applications. The proposed technique utilises the concept of minimum risk based classification to increase the bit depth of video frames and edge detection is used for frame matching to reduce the time complexity of the algorithm. The standard 8 bit depth video frame is downstream the low bit depth frames. The consecutive low bit depth frames are matchedusing edge detection. Minimum risk based classification is applied when the frames are notmatched and if the frames are matched the frame is replaced with the previous expandedframe. Using this method minimum risk based classification is not applied to every frameand the algorithm takes less time to simulate. The expanded images results in contouring effects in smooth regions having highly saturatedpixels. To reduce these effects image enhancement techniques are discussed in the thesis.These techniques involve discrete wavelet transform (DWT), color space transformation and simple linear iterative clustering (SLIC) segmentation. Simulation results reveal that the proposed technique is more time efficient and accurate than the other state of art video bit depth expansion techniques. The proposed method results in improved subjective and objective image quality with reduced time complexity.