【摘要】：智能手機(jī)在日常生活中的使用已經(jīng)相當(dāng)普遍,一些高端智能手機(jī)的相機(jī)像素也已經(jīng)達(dá)到甚至超過(guò)了數(shù)碼相機(jī)的像素。隨著軟件智能化和硬件微型化水平的不斷發(fā)展,人們對(duì)智能手機(jī)拍照功能的要求也日益提高。在這種情況下,特殊場(chǎng)景如星空、閃電和煙花等場(chǎng)景的拍照模式成為了眾廠商在智能手機(jī)攝像上研究的重點(diǎn)方向,也是競(jìng)爭(zhēng)相當(dāng)激烈的一個(gè)新興的發(fā)展領(lǐng)域。傳統(tǒng)的星空拍攝采用單次長(zhǎng)曝光法,但是受到電池電量和熱噪的影響,已經(jīng)逐步被疊加法取代,該方法采用較高的感光度,較大的光圈,較長(zhǎng)的曝光時(shí)間,連續(xù)拍攝多張照片,然后用后處理算法進(jìn)行疊加合成絢麗的星空。因此,如何克服高感光度帶來(lái)的噪聲,較長(zhǎng)曝光時(shí)間引入的模糊,達(dá)到完美的多幀合成效果都是手機(jī)拍攝星空所面臨的問(wèn)題。本文根據(jù)現(xiàn)今的手機(jī)對(duì)焦策略,采用基于圖像的自動(dòng)對(duì)焦方法,利用拉普拉斯算子作為評(píng)價(jià)函數(shù),極點(diǎn)搜索法作為搜索策略進(jìn)行自動(dòng)對(duì)焦。在拍照過(guò)程中,針對(duì)相機(jī)和星空?qǐng)鼍暗南鄬?duì)運(yùn)動(dòng)而導(dǎo)致的圖像模糊問(wèn)題,采用了一種新的針對(duì)星空?qǐng)D像的運(yùn)動(dòng)模糊復(fù)原算法。首先采用一種改進(jìn)的細(xì)化算法得到星空?qǐng)D像的星點(diǎn)運(yùn)動(dòng)軌跡,從而提取出星空運(yùn)動(dòng)模糊圖像的模糊核,然后基于Richardson-Lucy的復(fù)原算法對(duì)模糊圖像進(jìn)行復(fù)原。實(shí)驗(yàn)結(jié)果表明,本文采用的利用細(xì)化算法得到模糊核的方法復(fù)雜度低,耗時(shí)少且準(zhǔn)確性高,對(duì)于星空?qǐng)D像的運(yùn)動(dòng)模糊具有很好的復(fù)原效果。光照不足的條件下,針對(duì)開(kāi)啟高感光度引入的高感光度噪聲問(wèn)題,本文采用基于均值加速的快速中值濾波對(duì)星空?qǐng)D像進(jìn)行去噪,實(shí)驗(yàn)結(jié)果表明這種方法不僅時(shí)間復(fù)雜度低,還能有效的去除星空?qǐng)D像中的高感光度噪聲。為了能夠捕捉到不同狀態(tài)下,如亮暗不同、閃爍變化的星點(diǎn),采用基于特征點(diǎn)的方法對(duì)星空?qǐng)D像進(jìn)行了圖像配準(zhǔn),在星空?qǐng)D像配準(zhǔn)的前提下,根據(jù)星點(diǎn)的特點(diǎn),利用像素值取大的融合方法對(duì)星空?qǐng)鼍澳M圖像進(jìn)行多幀融合。該方法復(fù)雜度較低且效果較好,最終高效的獲得了完美的多幀融合后星空?qǐng)D像。針對(duì)手機(jī)在星空?qǐng)鼍芭臄z中出現(xiàn)的問(wèn)題,本文旨在提出一種合理的解決方案,并在實(shí)驗(yàn)室模擬實(shí)現(xiàn)相關(guān)算法。本文對(duì)于自動(dòng)對(duì)焦、手機(jī)抖動(dòng)導(dǎo)致的運(yùn)動(dòng)模糊復(fù)原、星空?qǐng)D像高感光度去噪以及星空?qǐng)D像的配準(zhǔn)融合四個(gè)方面的問(wèn)題進(jìn)行了深入研究以及不同算法的對(duì)比實(shí)驗(yàn),模擬星空?qǐng)鼍芭臄z的實(shí)驗(yàn)結(jié)果表明,經(jīng)過(guò)本文一系列星空?qǐng)鼍八惴ǖ奶幚?最終能夠獲得清晰完美的星空?qǐng)D像。
[Abstract]:The use of smart phones in daily life is already quite common, and some high-end smartphones have camera pixels that have reached or even exceeded those of digital cameras. With the development of software intelligence and hardware miniaturization, the requirement of smart phone photo-taking is increasing. In this case, the photography mode of special scenes such as starry sky, lightning and fireworks has become the focus of research on smartphone camera, and it is also a new field of development in which the competition is very fierce. Traditional star photography uses single long exposure method, but it has been replaced by superposition method because of the influence of battery power and thermal noise. This method uses higher sensitivity, larger aperture, longer exposure time, and takes many pictures continuously. Then the post-processing algorithm is used to superposition and synthesize the brilliant star sky. Therefore, how to overcome the noise caused by high luminosity, the blur introduced by long exposure time, and achieve the perfect multi-frame synthesis effect are the problems that the mobile phone faces in shooting star sky. In this paper, according to the present mobile phone focusing strategy, the image based auto-focusing method is adopted, Laplace operator is used as the evaluation function, and the pole search method is used as the search strategy. In order to solve the problem of image blur caused by the relative motion of camera and star scene, a new motion blur restoration algorithm for star-sky images is proposed in this paper. Firstly, an improved thinning algorithm is used to obtain the star motion trajectory of the star image, and then the blur kernel of the star motion blur image is extracted, and then the blur image is restored based on the Richardson-Lucy algorithm. The experimental results show that the proposed thinning algorithm has the advantages of low complexity, less time consuming and high accuracy, and has a good restoration effect for the motion blur of star images. Under the condition of insufficient illumination, aiming at the problem of high sensitivity noise caused by opening high sensitivity, this paper uses a fast median filter based on mean acceleration to Denoise the star image. The experimental results show that this method is not only low in time complexity, but also less complex in time. It can also effectively remove the high luminosity noise in the sky image. In order to capture the star points in different states, such as different brightness and dark, the star image is registered based on the feature point method. Under the premise of the star image registration, according to the characteristics of the star point, The multi-frame fusion method of star scene simulation image is presented in this paper. This method has low complexity and good effect. Finally, the perfect multi-frame fused star image is obtained efficiently. Aiming at the problem of mobile phone shooting in star-sky scene, this paper proposes a reasonable solution and implements the algorithm in laboratory simulation. In this paper, four aspects of automatic focus, mobile phone jitter caused motion blur restoration, star image high sensitivity denoising and star image registration fusion are studied in depth and the contrast experiments of different algorithms are carried out. The experimental results of simulated star scene shooting show that through the processing of a series of star scene algorithms in this paper, a clear and perfect star sky image can be obtained.
【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類(lèi)號(hào)】：TP391.41;TB86

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 鄭玉珍,吳勇,倪旭翔;實(shí)時(shí)自動(dòng)對(duì)焦的研究[J];光電工程;2004年04期

，

本文編號(hào)：2423961