Contour Artifact Removal for Expanded HDR Content

High Dynamic Range (HDR) displays are being widely adopted in the consumer electronics market, pushing up the demand of HDR content. However, HDR contents converted from their Standard Dynamic Range (SDR) counterparts often suffer from the contour artifact phenomenon. To address this issue, we formulate a sensitive region aware weighted gradient minimization process to remove the contour artifact. In order to assign reasonable weights to the image gradients in the objective function, we use features composed of local smoothness and gradient directions to highlight the risk regions, and then propose an expansion ratio to precisely locate the artifact. Solving the objective function gives improved HDR images with the contour artifacts minimized and texture preserved. Experimental results demonstrate the effectiveness of our method by the measurement of the objective metrics and visual comparison.