Abstract: “Pill color is applied as one of important features for pill identification; however, most of pill color is white. It’s difficult to classify two similar white pills because pill’s color is sensitive to luminance intensity. To reduce its luminance-intensity effect, RGB color is converted in term of YUV component because when its luminance intensity is increased, only its Y value is increased but its U and V values are almost constant. Therefore, its U and V values can be directly applied for image recognition. In this report, we propose luminance compensation based on background shadow to compensate luminance-intensity effect in Y value. When its luminance intensity is increased, Y value of background is also increased but Y value of pill shadow is slightly decreased. Therefore, difference between Y value of background and its shadow is effectively applied for luminance compensation. Experimental results confirm an effectiveness of our proposed compensation method.”

Keywords: Pill Identification, Shadow background, Luminance compensation, Digital image recognition, YUV