A Method of Obtaining Skin Color Distribution Areas Based on Illumination Compensation

Abstract

Background/Objectives: The ultra-high-speed camera captures even the slightest changes in ambient lighting as an image. Therefore, research is needed to correct the effect of lighting that is non-uniformly included in the image.

Methods/Statistical analysis: This paper proposes an algorithm that corrects irregular lighting from high-speed color images continuously input with a slight time interval, and then obtains an exposed skin region, which is a region of interest, from the corrected image. In this study, the non-uniform lighting effect is first corrected using a frame blending technique. Then, a region of interest is obtained from the image by applying an elliptical skin model.

Findings: Experimental results show that the proposed approach corrects the illumination from the input image and then accurately acquires the region of interest. In this study, the performance of the proposed method of acquiring skin pixels using illumination correction of images input at high speed was quantitatively compared and measured in terms of accuracy. In this study, a measure described as the relative ratio between the number of skin regions correctly acquired in the received images and the number of skin regions originally included in all images was used. In the conventional approach, the process of correcting irregular lighting in images was not normally performed. In addition, since a fixedly generated skin color distribution model is used, it is difficult to accurately extract skin pixels. On the other hand, the algorithm proposed in this paper effectively corrects the uneven lighting effect reflected in the image through frame blending between successive images, and then acquires the region of interest, allowing it to obtain skin color regions more robustly than conventional approaches.

Improvements/Applications: The proposed algorithm is expected to be useful in various kinds of practical application programs related to image recognition such as face recognition, lighting correction and removal, video indexing, etc.