一個基於人類視覺之影像對比強化的新方法

Abstract

近幾年由於影像擷取裝置技術的快速發展，數位相機和照像手機被廣泛的使用在生活中。由於影像擷取裝置的物理限制和拍照時環境光源的影響，易造成影像品質低落。不適當的光源易導致影像曝光過度或曝光不足之情形。而不均勻的光線以及空氣粒子所導致的光散射，使影像的對比度降低。為了解決這些問題，傳統的影像強化技術由於過度的將對比度拉開，造成真實影像的失真，或者無法同時有效強化影像中感興趣的區域(臉部)和背景區域。 因此，本論文提出一種有效的真實影像強化方法，利用人類視覺系統之恰可知覺差(Just Noticeable Difference)作為對比拉開的限制條件，改善傳統直方圖均化法之缺點。針對人像照片，我們延伸此方法並結合使用膚色光線校正，使人像照片能有對比適度的背景以及令人滿意的膚色光線。

Due to the rapid development of image capturing device techniques in recent years, digital cameras and camera phones have been widely used in our life. However, the limitation of capturing devices and improper exposure light may lead to unpleasant images. Inappropriate illumination condition may lead to underexposed or overexposed effects. In addition, non-uniform illumination and optical scattering of light by atmosphere particles may reduce the image contrast. To solve these problems, conventional image enhancement techniques may either fail to produce satisfactory and undistorted image owing to adjust the contrast excessively, or cannot be improved appropriately both in regions of interest, especially faces, and in the background. Hence, our thesis proposes an effective algorithm for real-life image enhancement regarding the just-noticeable-difference (JND) of human visual system (HVS) as a contrast-stretching constraint to remove the shortcomings of histogram equalization. Besides, we improve this approach for face images by combining skin dependent exposure correction to produce appropriate contrast in background and satisfied illumination in skin regions.