摘 要: 針對有明顯色彩和亮度差異的圖片,在進行傳統的圖像拼接后,會產生肉眼可見的拼接縫,嚴重影響拼接效果。因此,提出一種針對存在色差的圖像拼接算法。首先,計算相鄰圖片間的色彩校正參數和每個彩色通道的全局色彩調整參數,選擇適合的顏色和亮度,對每一幅圖片進行色彩校正。其次,引入SIFT和RANSAC算法,實現了圖像的準確配準。最后,采取0?1加權融合算法和平均融合算法結合進行圖像融合,得到最終的全景圖像。實驗結果表明,由于已經進行圖像間的校正,減少了待拼接圖像的色差,使得圖像的融合更加簡單。該算法對于存在色差的圖片序列可以實現無縫快速的拼接,并且可根據需要調整圖像的色彩和亮度。
關鍵詞: 圖像拼接; 色彩校正; SIFT; 圖像融合
中圖分類號: TN911.73?34; TP391 文獻標識碼: A 文章編號: 1004?373X(2013)23?0057?04
Image stitching algorithm for images with color difference
WANG Li1, YI Zi?chuan2, PAN Zhong?liang1, BAI Peng?fei2
(1. School of Physics Telecommunication Engineering, South China Normal University, Guangzhou 510006, China;
2. South China Academy of Advanced Optoelectronics, South China Normal University, Guangzhou 510006, China)
Abstract: An image stitching algorithm is proposed for the images with color difference because the image stitching between two adjacent images with differences in colors and luminance could produce visible color difference and seams when the traditional algorithm is used, and the image stitching effect is affected seriously. The color correction coefficients between adjacent images and global color adjustment parameters of each color channel are calculated, and color correction for each image is carried out by selection of appropriate color and brightness. SIFT and RANSAC algorithms are introduced and an accurate image registration is realized. 0?1 weighted fusion algorithm and average fusion algorithm are combined to realize the image fusion, and a final panorama image is gotten. Experimental results show that the proposed algorithm can realize seamless fast stitching between images with color difference. In addition, the image fusion algorithm becomes easier than the traditional algorithm by reducing the color difference of images.
Keywords: image stitching; color correction; SIFT; image fusion
0 引 言
圖像拼接或者全景圖像技術是指把同一場景的相鄰之間有重合區域的一系列圖片拼接成一幅寬視角的、與原始圖像接近且失真小、沒有明顯縫合線的圖片,以滿足人們瀏覽全景圖片的需求。該技術正被應用于各種領域,如視頻壓縮、視頻索引、目標跟蹤[1]或者虛擬環境的創建。
目前,已經有許多算法都可以將同一場景圖片中的重疊部分提取出來并且快速的拼接成為一幅全景圖片,其中有許多算法可以得出令人滿意的融合結果。Ming?Shing Su 等使用小波進行多尺度的融合[2],但是這種方法只是在重疊區域進行局部的融合處理,無法保證圖像的全局自然。……