For The Comparative Study Of The Building Detection Bayer Interpolation Algorithm True Color Images

In recent years, the digital aerial camera in the aviation field is applied more and more widely. Compared with traditional analog film camera and satellite multiband sensors, the digital aerial camera has different imaging mode and image features. The imaging mode of digital aerial camera is Bayer true color. Bayer true color image get the red, green, blue(R、G、 B) three bands of the image by CFA (Color Filter Array, digital Filter arrays), and then using the image reconstruction algorithm to obtain more approach the true color of the ground to reconstruct ground remote sensing images. Image reconstruction algorithm for general digital aerial camera usually based on image quality and visual effects, we need to do some further research on the applicability of various image reconstruction algorithm applied in the aerial remote sensing applications to seek more suitable mode of image reconstruction for data processing and application in remote sensing, such as edge detection, image segmentation, target detection, etc. This paper is based on this idea to expand the correlation research.This paper starts from Bayer true color image imaging characteristics, according to the influencing factors of building detection analysis of selection of color restoration algorithm. Using nearest neighbor, bilinear method, based on the gradient interpolation method and the adaptive interpolation method to restore Bayer true color image respectively. Compared with the four kinds of interpolation results by traditional visual interpretation method, that the four interpolation algorithms all can restore image information very well, but there is still exist some slight difference on noisy point and color distortion problems. Base on this result we can rank these four interpolation algorithm as adaptive interpolation, based on gradient interpolation, bilinear interpolation and nearest neighbor interpolation.In the part of algorithm evaluation, this paper uses building detection method as the breakthrough point and simulates the experiment process of building detection by simulation experiment, using analytic hierarchy process to evaluate the results. This paper uses three kinds of evaluation methods:the first one is the RMSE and MSSIM parameters values for quality evaluation, by means of mathematics in quantitative evaluation of color restoration results map distortion measure, and then determine the correlation between images. The experimental data show that based on the gradient interpolation method has better effect on the fidelity, the comprehensive weighted values higher than other algorithms greatly. Secondly, edge detection method. Based on the geometrical characteristic of buildings on the continuity of image to determine, using the Canny operator, Sobel operator and Roberts operator to detect the image edge, and then calculate the MSSIM values between images. According to the MSSIM value of each algorithm assigned index value, by comparing each algorithm index value, nearest neighbor interpolation is better than others. The third one is Image segmentation method for building detection, image segmentation and extraction base on texture feature. Comparison of visual interpretation and parameters to judge the results, the interpolation algorithm based on gradient is the best. Finally, all the results will be integrated again by AHP(analytic hierarchy process), and verification experiments operated with a large amount of data, the most suitable interpolation algorithm for Bayer true color image is the interpolation algorithm based on gradient.