基于Color Lines先驗(yàn)的高階馬爾科夫隨機(jī)場去霧
doi: 10.11999/JEIT151308 cstr: 32379.14.JEIT151308
基金項(xiàng)目:
國家自然科學(xué)基金(61372167, 61379140)
Higher-order Markov Random Fields Defogging Based on Color Lines
Funds:
The National Natural Science Foundation of China (61372167, 61379140)
-
摘要: 傳統(tǒng)的一階馬爾科夫隨機(jī)場在圖像先驗(yàn)信息表達(dá)和對圖像整體的約束上能力有限，同時(shí)基于暗通道的去霧算法在天空等大片白色區(qū)域處理效果存在偏差。針對以上問題，該文提出一種基于Color Lines 的高階馬爾科夫隨機(jī)場去霧算法。該算法通過引入對顏色失真具有很好魯棒性的Color Lines 先驗(yàn)條件，初步校正經(jīng)暗通道獲取的傳輸圖，然后利用高階馬爾科夫隨機(jī)場優(yōu)化傳輸圖，獲取最終精確的去霧圖像。實(shí)驗(yàn)結(jié)果表明，與已有算法相比，該文算法具有更強(qiáng)的普適性，可提高霧天圖像的清晰度，同時(shí)恢復(fù)更多的圖像細(xì)節(jié)等信息。
-
關(guān)鍵詞:
- 圖像去霧 /
- 暗通道先驗(yàn) /
- 高階馬爾科夫隨機(jī)場 /
- Color Lines
Abstract: Compared with the first-order Markov random fields, higher-order Markov random fields could incorporate more statistical priors, thus have much expressive power of modeling. And the defogged images which based on dark channel prior have much error in sky regions and big white blocks. To solve those problems, this paper proposes a Markov random fields defogging method based on Color Lines. This method corrects the dark channel prior, according to the color lines which has a good robustness to color distortion, then uses the higher-order Markov random fields to optimize the transmission image to obtain final defogged image. The experimental results show that this method could improve the image resolution, while maintaining more image details.-
Key words:
- Image defogging /
- Dark channel prior /
- Higher-order Markov random fields /
- Color Lines
-
周妍, 李慶武, 霍冠英. 基于非下采樣Contourlet變換系數(shù)直方圖匹配的自適應(yīng)圖像增強(qiáng)[J]. 光學(xué)精密工程, 2014, 22(8): 2214-2222. doi: 10.3788/OPE.20142208.2214. ZHOU Yan, LI Qingwu, and HUO Guanying. Adaptive image enhancement based on NSCT coefficient histogram matching [J]. Optics and Precision Engineering, 2014, 22(8): 2214-2222. doi: 10.3788/OPE.20142208.2214. SEOW M J and ASARI V K. Ratio rule and homomorphic filter for enhancement of digital color image[J]. Neurocomputing, 2006, 69(7/9): 954-958. doi: 10.1016 /j.neucom.2005.07.003. 趙宏宇, 蕭創(chuàng)柏, 禹晶. 馬爾科夫隨機(jī)場模型下的Retinex夜間彩色圖像增強(qiáng)[J]. 光學(xué)精密工程, 2014, 22(4): 1048-1055. doi: 10.3788/OPE.20142204.1048. ZHAO Hongyu, XIAO Chuangbo, and YU Jing. A Retinex algorithm for night color image enhancement by MRF[J]. Optics and Precision Engineering, 2014, 22(4): 1048-1055. doi: 10.3788/OPE.20142204.1048. NARASIMHAN S G and NAYAR S K. Chromatic frame-work for vision in bad weather[C]. IEEE Conference on Computer Vision Pattern Recognition, South Carolina, USA, 2000, 1: 598-605. NARASIMHAN S G and NAYAR S K. Interactive (de) weathering of an image using physical models[C]. IEEE Workshop on Color and Photometric Methods in Computer Vision, Nice, France, 2003: 1-8. TAN R T. Visibility in bad weather from a single image[C]. IEEE Conference on Computer Vision and Pattern Recognition, Anchorage, AK, USA, 2008: 1-8. doi: 10.1109/ CVPR.2008.4587643. FATTAL R. Single image dehazing[J]. ACM Transactions on Graphics, 2008, 27(3): 1-9. doi: 10.1145/1360612.1360671. HE K, SUN J, and TANG X. Single image haze removal using dark channel prior[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(12): 2341-2353. doi: 10.1109/ TPAMI.2010.168. 余淼, 胡占義. 高階馬爾科夫隨機(jī)場及其在場景理解中的應(yīng)用[J]. 自動化學(xué)報(bào), 2015, 41(7): 1213-1234. doi: 10.16383/ j.aas.2015.c140684. YU Miao and HU Zhanyi. Higher-order Markov random fields and their applications in scene understanding[J]. Acta Automatica Sinica, 2015, 41(7): 1213-1234. doi: 10.16383/ j.aas.2015.c140684. FATTAL R. Dehazing using color-lines[J]. ACM Transactions on Graphics, 2014, 34(1): 1-14. doi: 10.1145/2651362. CHAO S and TSAI D. An improved anisotropic diffusion model for detail and edges-preserving smoothing[J]. Pattern Recognition Letters, 2010, 31(13): 2012-2023. doi: 10.1016/ j.patrec.2010.06.004. ANDREW B, PUSHMEET K, and Carsten R. Markov Random Fields for Vision and Image Processing[M]. Massachusetts London, England, The MIT Press, Cambridge, 2011: 311-328. -
計(jì)量
- 文章訪問數(shù): 1683
- HTML全文瀏覽量: 188
- PDF下載量: 531
- 被引次數(shù): 0
下載:
