Ultrasound video despeckle filtering for High Efficiency Video Coding in m-health systems

Abstract

Mobile-Health (m-health) medical video communication systems need to communicate ultrasound video of adequate diagnostic quality for remote diagnosis and care. In this study, we demonstrate how ultrasound video noise reduction, employed as a pre-processing step, used in conjunction with High Efficiency Video Coding (HEVC), can significantly improve diagnostic quality at considerably lower bitrate demands than conventional H.264/AVC source encoding. Three video despeckle filters were investigated, namely: linear filtering (DsFlsmv), hybrid median filtering (DsFhmedian), and speckle reducing anisotropic diffusion (DsFsrad). The methods were tested using ten atherosclerotic plaque ultrasound videos. Speckle filtering was applied at a frame-level basis. Experimental results were obtained using both subjective (by two clinical experts) and objective video quality assessment. The best achieving despeckle filter was DsFlsmv followed closely by DsFhmedian. In both cases, the filters provided for increased clinical performance while associated with significant bitrate demands reductions.