在本論文中,我們探討增強低動態範圍影像細節的方法。因為人眼具有高動態範圍的辨識能力,所以可以辨識出低動態範圍影像的細節,但是當影像在低動態範圍的設備上輸出,許多細節就難以辨認,然而在許多場合中,只能透過低動態範圍的設備觀察或是監看場景,因此,在低動態範圍的輸出設備上輸出近似人眼看到的實際影像,成為一個蓬勃發展的研究課題。本論文使用區域適應性直方圖修正法以增強低動態範圍影像,輸入影像切割成不重疊的區塊,計算其直方圖,並據以計算影像輸入值和輸出值的對應曲線。為了避免亮度集中區域的亮度被強制拉開,產生過度強化的現象,我們對於亮度對比也加上了限制。為了避免產生區塊效應,像素最後的亮度是周圍亮度的平均值。對於彩色影像,我們先將其轉換成YUV表示法,對於Y值進行影像增強,最後將其轉換回原始影像的色彩空間。我們的方法可以增強影像亮部與暗部的對比度,增加細節的辨識度,又可以有效地維持影像亮度適中的部份,使得低對比影像經過強化之後,在低動態範圍設備上有良好的顯示結果。
In this thesis, we investigate methods of enhancing the low dynamic range image and propose a method of adaptive image contrast enhancement. Because human eyes have high dynamic range, we can recognize details from scenes. When a scene is captured by devices such as digital camera and displayed on the low dynamic range device, it may be hard to recognize details. However, under many situations, we can only watch or monitor scenes on the low dynamic range device. Therefore, it is important to display images with enhanced details on low dynamic range device. We propose a local adaptive histogram modification method to enhance low dynamic range images. In each non-overlapping image block, we calculate the intensity mapping curve from the histogram of the block. To avoid over-enhancement, we limit the contrast by clipping the histogram values. To prevent block effects, the final pixel intensity comes from the average of the corresponding intensities of the surrounding blocks. For color images, we convert the pixel color values to equivalent YUV values and do enhancement in Y domain. Then convert the Y-enhanced image back to the one in the original color space. From our experiments, our method can enhance the image contrast on either bright or dark regions, thereby improving image details. Furthermore, our algorithm keeps the image nearly unchanged on the regions of moderate brightness. The images processed by the proposed algorithm display well in terms of contrast on the low dynamic range device.