單幅霧天圖像去霧算法研究
發(fā)布時間：2018-12-13 09:44
【摘要】：近年來,霧霾一次又一次頻繁的出現(xiàn)在大家的視野之內(nèi),不僅影響到人們的身體健康,同時嚴重危害到人們的公共交通安全。針對霧天天氣情況下獲取的各種圖像進行去霧研究,本文分析了幾種基于圖像增強的去霧算法在去霧領(lǐng)域的不足之處。同時,在前人對基于模型的去霧的研究以及在進行理論分析的基礎(chǔ)之上,提出透射率趨勢圖去霧的概念,并針對傳統(tǒng)的暗原色先驗理論去霧存在的問題提出幾點改進想法。(1)針對傳統(tǒng)暗原色明亮區(qū)域恢復后顏色失真問題,提出明亮區(qū)域透射率補償方法進行修正。通過引入明亮區(qū)判別閾值,然后根據(jù)原始圖像像素亮度值與大氣光值差的絕對值與閾值的大小作為判別依據(jù),比閾值小則為明亮區(qū),然后通過透射率修正函數(shù)進行修正;反之為非明亮區(qū),則保持原有的透射率。實驗證明本文所提透射率修正方法有效解決了明亮區(qū)顏色失真現(xiàn)象;(2)針對傳統(tǒng)暗原色恢復結(jié)果亮度偏暗以及過飽和現(xiàn)象。通過分析大氣散射理論并結(jié)合成像設(shè)備的局限性提出采用透射率趨勢圖進行去霧的概念。文中采用對粗透射率圖進行空間平滑濾波的思想來獲取透射率趨勢圖。實驗證明,利用透射率趨勢圖具有較好的去霧效果,改善了原始算法去霧結(jié)果亮度偏暗以及過度飽和的問題;(3)針對原始算法效率低下的問題,本文結(jié)合透射率趨勢圖的概念,提出兩種粗透射率圖濾波平滑的改進方法:其一,基于高斯平滑濾波的概念,提出基于改進的高斯空間平滑濾波處理來獲取透射率趨勢圖;其二,基于改進的引導濾波的方法,通過重構(gòu)引導矩陣對粗透射率圖平滑處理,不僅能夠獲得等價細膩的透射率趨勢圖,同時降低了傳統(tǒng)引導濾波對透射率優(yōu)化的時間開銷,其算法的執(zhí)行效率為原始算法的3~4倍;(4)基于可見邊增強的去霧質(zhì)量評價方法,通過引入色彩信息恢復度指標,結(jié)合清晰度指標,構(gòu)建無參考的圖像去霧質(zhì)量綜合評價指標,更加全面的對去霧結(jié)果進行評價,論文基于Matlab平臺,編寫相關(guān)算法程序,對上述算法進行實驗仿真。實驗結(jié)果表明,本文提出的改進方案有效的改善和解決了傳統(tǒng)暗原色先驗去霧算法存在的問題。
[Abstract]:In recent years, haze appears frequently in the field of vision, which not only affects people's health, but also seriously endangers people's public transportation safety. In this paper, we analyze the shortcomings of several image enhancement algorithms in the field of fog removal. At the same time, on the basis of previous researches on model-based de-fogging and theoretical analysis, the concept of transmittance trend map de-fogging is proposed. In order to solve the problem of removing fog in the traditional priori theory of dark primary color, some improvement ideas are put forward. (1) aiming at the problem of color distortion after restoration of the bright region of traditional dark primary color, a method of compensation for the transmittance of bright region is proposed. By introducing the bright region to distinguish the threshold value, then according to the original image pixel brightness value and atmospheric light value difference between the absolute value and the size of the threshold as the basis, smaller than the threshold value is the bright region, and then through the transmittance correction function to correct; On the contrary, if the region is not bright, the original transmittance will be maintained. Experiments show that the proposed transmittance correction method can effectively solve the bright color distortion phenomenon. (2) aiming at the traditional dark primary color restoration results brightness dim and supersaturation phenomenon. Based on the analysis of atmospheric scattering theory and the limitation of imaging equipment, the concept of defogging using transmittance trend map is proposed. In this paper, the idea of spatial smoothing filter for coarse transmittance map is used to obtain the transmittance trend map. Experimental results show that the transmittance trend map has a better effect on fog removal and improves the problem of dark brightness and over-saturation of the original algorithm. (3) aiming at the problem of low efficiency of the original algorithm, this paper proposes two improved methods of filtering smoothing based on the concept of transmittance trend map: first, based on the concept of Gao Si smoothing filter, Based on the improved Gao Si spatial smoothing filter, the transmittance trend map is obtained. Secondly, based on the improved guided filtering method, by reconstructing the guide matrix to smooth the coarse transmittance map, not only the equivalent and delicate transmittance trend map can be obtained, but also the time cost of the traditional guided filter to optimize the transmittance can be reduced. The efficiency of the algorithm is 3 times that of the original algorithm. (4) based on the evaluation method of visible edge enhancement, by introducing the restoration index of color information and combining the definition index, the comprehensive evaluation index of image de-fogging quality without reference is constructed to evaluate the result of de-fogging more comprehensively. Based on the Matlab platform, this paper compiles the related algorithm program, and simulates the above algorithm. The experimental results show that the proposed scheme can effectively improve and solve the problems of the traditional dark color priori de-fogging algorithm.
【學位授予單位】：東華理工大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP391.41
[Abstract]:In recent years, haze appears frequently in the field of vision, which not only affects people's health, but also seriously endangers people's public transportation safety. In this paper, we analyze the shortcomings of several image enhancement algorithms in the field of fog removal. At the same time, on the basis of previous researches on model-based de-fogging and theoretical analysis, the concept of transmittance trend map de-fogging is proposed. In order to solve the problem of removing fog in the traditional priori theory of dark primary color, some improvement ideas are put forward. (1) aiming at the problem of color distortion after restoration of the bright region of traditional dark primary color, a method of compensation for the transmittance of bright region is proposed. By introducing the bright region to distinguish the threshold value, then according to the original image pixel brightness value and atmospheric light value difference between the absolute value and the size of the threshold as the basis, smaller than the threshold value is the bright region, and then through the transmittance correction function to correct; On the contrary, if the region is not bright, the original transmittance will be maintained. Experiments show that the proposed transmittance correction method can effectively solve the bright color distortion phenomenon. (2) aiming at the traditional dark primary color restoration results brightness dim and supersaturation phenomenon. Based on the analysis of atmospheric scattering theory and the limitation of imaging equipment, the concept of defogging using transmittance trend map is proposed. In this paper, the idea of spatial smoothing filter for coarse transmittance map is used to obtain the transmittance trend map. Experimental results show that the transmittance trend map has a better effect on fog removal and improves the problem of dark brightness and over-saturation of the original algorithm. (3) aiming at the problem of low efficiency of the original algorithm, this paper proposes two improved methods of filtering smoothing based on the concept of transmittance trend map: first, based on the concept of Gao Si smoothing filter, Based on the improved Gao Si spatial smoothing filter, the transmittance trend map is obtained. Secondly, based on the improved guided filtering method, by reconstructing the guide matrix to smooth the coarse transmittance map, not only the equivalent and delicate transmittance trend map can be obtained, but also the time cost of the traditional guided filter to optimize the transmittance can be reduced. The efficiency of the algorithm is 3 times that of the original algorithm. (4) based on the evaluation method of visible edge enhancement, by introducing the restoration index of color information and combining the definition index, the comprehensive evaluation index of image de-fogging quality without reference is constructed to evaluate the result of de-fogging more comprehensively. Based on the Matlab platform, this paper compiles the related algorithm program, and simulates the above algorithm. The experimental results show that the proposed scheme can effectively improve and solve the problems of the traditional dark color priori de-fogging algorithm.
【學位授予單位】：東華理工大學
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：TP391.41
【參考文獻】
相關(guān)期刊論文 前10條
1 邢曉敏;劉威;;霧天交通場景中單幅圖像去霧[J];中國圖象圖形學報;2016年11期
2 龔昌來;羅聰;;一種改進的容差機制圖像去霧算法[J];液晶與顯示;2016年11期
3 紀紅霞;姚軍;潘延鵬;舒?zhèn)?;基于暗通道先驗的無人機圖像去霧算法研究[J];光學與光電技術(shù);2016年05期
4 薛玉利;;霧霾天氣情況下的交通標志檢測[J];交通運輸系統(tǒng)工程與信息;2016年04期
5 甘玉泉;汶德勝;王樂;高曉惠;衛(wèi)翠玉;;暗通道自然災害遙感圖像去霧[J];光子學報;2015年06期
6 吳迪;朱青松;;圖像去霧的最新研究進展[J];自動化學報;2015年02期
7 汪榮貴;傅劍峰;楊志學;沈法琳;查煒;;基于暗原色先驗模型的Retinex算法[J];電子學報;2013年06期
8 褚宏莉;李元祥;周則明;沈霽;;基于黑色通道的圖像快速去霧優(yōu)化算法[J];電子學報;2013年04期
9 劉楠;程詠梅;趙永強;;基于加權(quán)暗通道的圖像去霧方法[J];光子學報;2012年03期
10 黃黎紅;;單幅圖像的去霧新算法[J];光子學報;2011年09期
相關(guān)博士學位論文 前4條
1 羅學剛;面向災害應急響應的無人機影像快速處理技術(shù)研究[D];成都理工大學;2015年
2 嵇曉強;圖像快速去霧與清晰度恢復技術(shù)研究[D];中國科學院研究生院（長春光學精密機械與物理研究所）;2012年
3 郭t，
本文編號：2376335
本文鏈接：http://sikaile.net/kejilunwen/ruanjiangongchenglunwen/2376335.html
最近更新
教材專著
