数码相机的自动白平衡算法研究及实现

发布时间：2018-02-09 06:19

本文关键词： 自动白平衡色温灰度世界法完美反射法增益 NT96211　出处：《湖南科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：数码相机自诞生以来，3A(自动聚焦、自动曝光、自动白平衡)算法一直是被研究的核心内容，，本文主要是对自动白平衡算法进行研究及实现。由于人类视觉具有颜色恒常性，对于在不同光源下的同一物体感知为相对恒定的颜色。而数码相机在低色温光源下拍摄的照片偏红，在高色温光源下拍摄的照片偏蓝，针对这一现象，相机必须进行白平衡调节。所谓白平衡从字面意思可以理解为对白色进行调节使之到达平衡，不管在何种色温的光源下拍摄白色的物体，都能将白色的物体还原成为原本的白色。在研究现有的自动白平衡算法的前提下，为了克服灰度世界法在单一色彩的场景白平衡失效，以及图像中有亮度很高像素存在时完美反射法失效的缺陷，本文提出了一种基于光源色温判断的自动白平衡算法。首先，方便计算和统计光源的属性，将整幅图像分为1616个宏块。然后，根据提供的各种色温下光源的R/G、B/G的范围来确定每一个宏块属于哪种光源。如果超过50个宏块属于某种光源，则将此种光源记为100分，其他情况则根据函数关系得分。单一依靠R/G、B/G的范围来确定光源属性会出现判错的情况，光源错判主要以亮度为标准评价，对于光源错判有相应的减分原则。通过最后得分结果判断整幅图像的所属光源，分数最高的光源即为整幅图像的光源。最后，计算白平衡增益。根据灰色法原理，在某种确定光源下白点的红、绿、蓝三基色的灰度值相等。将绿色通道的增益值确定为256，计算Rgain、Bgain的白平衡增益。为了评估算法的白平衡调整效果，将该算法移植到NT96211搭建的硬件平台进行验证。采用了主观和客观两种方式来评估本文提出的白平衡算法。主观方式主要是人眼观察图片，而客观方法采用图像处理软件采集到的数据进行比较客观地分析白平衡算法的效果。两种评估方式都表明，本文提出的白平衡算法相对于典型的算法有很大的改善。同时该算法与一般数码相机白平衡算法最大的区别在于：它的白平衡参数是依据分析当前帧的图像信息得到的，而一般数码相机白平衡算法的白平衡参数是依据拍照前最后一帧图像的信息计算得到，故该优势更适合在红外监控类产品中使用。
[Abstract]:Since the digital camera was born, the algorithm of automatic focusing, automatic exposure and automatic white balance has always been the core of the research. This paper mainly studies and realizes the algorithm of automatic white balance. Because of the color constancy of human vision, In view of the fact that the same object under different light sources is perceived as a relatively constant color, and that the photos taken by a digital camera under a low color temperature light source tend to be red, and the photos taken under a high color temperature light source tend to be blue, The camera must adjust the white balance. The so-called white balance is literally understood as adjusting the white to balance, regardless of the color temperature under which white objects are photographed. Can restore the white object to the original white. On the premise of studying the existing automatic white balance algorithm, in order to overcome the failure of the gray world method in the single color scene white balance, and the defect of perfect reflection method when there are very high brightness pixels in the image, In this paper, an automatic white balance algorithm based on color temperature judgement of light source is proposed. Firstly, it is convenient to calculate and calculate the properties of light source, and the whole image is divided into 166 macroblocks. Determine what kind of light source each macroblock belongs to based on the range of R / G / B / G of the light source provided at various color temperatures. If more than 50 macroblocks belong to a certain light source, mark the light source at 100 points. In other cases, according to the function relation, the light source attribute will be judged wrong by relying solely on the range of R / G / G / B / G, and the light source error judgment is mainly based on the brightness as the standard evaluation. According to the result of the final score, the light source with the highest score is the light source of the whole image. Finally, the white balance gain is calculated. The gray values of red, green and blue primary colors are equal under a certain light source. The gain value of green channel is determined to be 256, and the white balance gain of RgainBgain is calculated. In order to evaluate the white balance adjustment effect of the algorithm, the algorithm is transplanted to the hardware platform set up by NT96211 for verification. The white balance algorithm proposed in this paper is evaluated by subjective and objective methods. The objective method uses the data collected by the image processing software to analyze the effect of the white balance algorithm objectively. The white balance algorithm proposed in this paper is much better than the typical one. At the same time, the biggest difference between the white balance algorithm and the general white balance algorithm is that the white balance parameters are obtained by analyzing the image information of the current frame. The white balance parameters of the general white balance algorithm of digital cameras are calculated according to the information of the last frame of image before taking pictures, so this advantage is more suitable for use in infrared monitoring products.
【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.41;TB852.1

【参考文献】