Evaluation And Optimization Of The Image Color In Surveillancee Video

Video surveillance system is widely used in the security field in recent years.It is vitally important to reproduce the image color (of videos) accurately for observation and obtaining evidence. The human eyes have the characteristic of color constancy, which ensures that the color observed is not affected by the color temperature of the light source. However, due to the lack of adaption, machine vision doesn’t have the same characteristic. In addition to the different colors of light source itslf, the performance of image acquisition devices and image display devices can also affect color reproduction of images..In order to eliminate image color distortion in the surveillance videos, this thesis uses appropriate ways to evaluate and optimize the image color, so as to realize color reproduction.Among all the steps of image tuning, the automatic white balance algorithm is the key step. After researching existing White balance algorithms, the thesis puts forward an automatic white balance algorithm based on color temperature estimation. It not only effectively overcomes the failure of the gray world algorithm when images contain too few colors, but also remarkably reduces the computational complexity comparing with neural network algorithm. The white balance algorithm introduced in this thesis mainly has four steps, first of all, discovering the distribution law of gray point of Cb, Cr in common lights through the experiment. Then according to the curve, interpolating out other color temperature value points and determining the corresponding color temperature distribution range of gray point of Cb, Cr. Next calculating the Cb and Cr values in image pixels to estimate the color temperature of the light source in the image. Finally calculating the corresponding gains, to adjust the image.The thesis introduces two kinds of evaluation method of image color, the subjective evaluation and objective evaluation, respectively. Chromatic aberration method is used here to evaluate image color optimization results. After applying the automatic white balance algorithm to image color correction, image color reproduction has been basically achieved. But there are still some deficiencies according to the evaluation result. So the thesis further optimizes the image color by correcting the saturation and brightness of colors. This time, the evaluation result meets the established requirements, the total chromatic aberration △E*ab is less than 15, the chromaticity aberration △C*ab is less than 10, and color saturation value is between 100 and 130. Finally, the thesis sets three color models, which are natural, standard, and bright-colored, to meet diverse needs of customers.