以醫美實例為基礎的去斑與美白

Abstract

最近，藉由雷射療程來去斑與美白已經成為了流行，在本文中我們提出一個分析雷射美白療程前後的照片中的黑色素與血紅素來去斑與轉換美白效果的系統。對於去斑與美白轉換，首先我們會分別建立去斑資料集與美白資料集，並對當中每對療程前後的照片分析其中的黑色素與血紅素，而對於待處理的照片，我們分析其中的黑色素與血紅素並偵測照片中斑點的位置。接著，我們由去斑與美白資料集中尋找適合的資料並將其去斑與美白效果套用至待處理照片中。藉由我們提出的系統對待處理的照片進行去斑與美白效果的轉換，我們的結果與實際做完雷射療程的結果相似並且與市面上只為追求漂亮的美白軟體相比我們的結果更為自然。

Using laser treatment to retouch the freckle and whiten the skin has become very popular recently. In this thesis, we propose an example-based system for freckle retouching and whitening effect transfer based on an example set of image pairs obtained before and after the laser cosmetic treatment. The proposed freckle retouching and whitening effect transfer are based on the melanin and hemoglobin extracted from face images. For freckle retouching and whitening effect transfer, we first build a retouch example set and whitening example set, and then extract the melanin and hemoglobin for each image pair. For a query image, we extract the melanin and hemoglobin and locate the freckle regions. The appropriate sets from retouch example set and whitening example set are found and used to do the freckle retouching and whitening effect transfer. Using the proposed system, for a query image we can do the freckle retouching and whitening effect transfer, resulting in an image that looks close to the result of laser treatment and more natural than the image beautified by using whitening apps available on the market.