Research And Implementation Of Industrial X-ray Image Enhancement Algorithms Based On CUDA

Today X-ray inspection has become one of the conventional techniques of nondestructive testing,improve the quality of X-ray images has always been the development direction of non-destructive testing technology.Through the digital image processing technology to improve the image with low signal to noise ratio,low illumination,low contrast and other problems and enhancing image details ? improving image quality is the most important method to improve the detection effect.For industrial ray images,traditional digital image enhancement algorithms can not meet the needs of real time processing in practical applications because of the large amount of computing data and complex calculation methods.The emergence of CUDA has designed a new architecture for GPU computing,thus solving the limitations of GPU for image enhancement.Therefore,image enhancement algorithm based on CUDA has great advantages over traditional algorithms in speed.The main work is organized as follows:1.The development of X-ray and image enhancement technology at home and abroad is briefly described.The GPU acceleration technology of CUDA architecture is analyzed from several levels,such as hardware organization,thread structure,programming model and so on.2.A classical radiographic image enhancement algorithm,MSR algorithm,is analyzed in depth,MSR algorithm has better enhancement effect on radiographic images,but its parameters are difficult to choose and its adaptability to images is poor.The local standard deviation and image gradient are introduced to measure the local features of the image,adaptively select the scale factor of the algorithm and the weight of each scale,this improvement not only improves the adaptability of the image,but also improves the effect ofthe algorithm.Finally,according to the characteristics of the algorithm,the parallelism is optimized based on the CUDA architecture,so that the algorithm can meet the real-time processing requirements in practical applications.3.The gradient field enhancement algorithm has a good effect on the ray image processing,but the overall visual effect of the enhanced image is poor,and the image integral and local information can not be taken into account.Therefore,the perceptual characteristics of the human eye are introduced to improve the image.Firstly,the minimum perceptible difference of the image is introduced according to the luminance masking characteristics,and the average brightness of the scene background is calculated after the normalization of the image,and the minimum detectable difference is obtained by combining the image gradient.According to the minimum perceptible difference,the image is divided into four luminance regions,and the image gradient is adjusted respectively to the four regions,and the enhanced image is reconstructed from the target by solving the Poisson equation.The algorithm is optimized based on CUDA architecture,including parallel computation of image gradient and parallel solution of Poisson equation.