Quality Guided Skin Tone Enhancement for Portrait Photography

Quality-Guided Skin Tone Enhancement for Portrait Photography

Abstract:

Portrait photography plays a vital role in personal and professional communication, where skin tone quality significantly affects the visual appeal of an image. However, factors such as lighting conditions, camera settings, and color space limitations often result in inaccurate or unpleasant skin tone reproduction. This paper presents a Quality-Guided Skin Tone Enhancement system for portrait photography that intelligently improves skin tone appearance using artificial intelligence and image processing techniques. The proposed method converts portrait images from the RGB color space to the perceptually uniform CIELAB color space to enable accurate color manipulation. K-means clustering is applied to group images with similar skin tone characteristics into meaningful clusters. A convolutional neural network–based feature extraction module captures image statistics, a quality score, and the skin tone cluster label to generate enhancement parameters. The system predicts both one-dimensional and three-dimensional Look-Up Tables to adjust luminance, color balance, and saturation in a controllable manner. Enhancement intensity is guided by a quality score ranging from low to high, giving users flexible control over the output. Quantitative evaluation using PSNR, SSIM, and Delta-E metrics demonstrates that the proposed method consistently outperforms the baseline identity LUT approach, producing visually pleasing and perceptually accurate portrait skin tones across diverse racial categories.

Keywords: Skin Tone Enhancement, Portrait Photography, Look-Up Table, CIELAB Color Space, Quality-Guided Image Processing.