The Study Of The Technology Of X-ray Image Dynamic Enhancement Based On The GPU Accelerating

With the acceleration of production rhythm,due to the external environment, industrial ray detection will cause the reduction of image quality, equipment and other factors in image generation and transmission, thus the detection need to process dynamic real-time enhance . At present, the dynamic enhanced systems have carried out by the use of DSP and FPGA acceleration, which can not only increase the cost of the system but also restrict on the specific hardware so as to not be easily modified. At the same time, the graphics processor unit (GPU) has the feature of high-speed in graphics pipeline and parallelism, and the developed programmable features in recent years have made great potential in general calculate area of the digital image processing. Response to the above, this article has studied the dynamic enhancement on ray image, using the GPU accelerating technology based on DirectX, in the condition of not increasing specific hardware.This article introduces the current commonly used methods of image processing hardware acceleration, in particular, introduces the GPU acceleration based on DirectX, and finally build GPU-accelerated ray image processing system on PC by using of DirectX technology. This system makes full use of the parallel architecture of graphics hardware, inputs the image in the form of texture to the chip element rendering program and carries out the same type of operation processing in all grid points by the way of parallel data streams, at the same time output the real-time results. In this system, for image of the stationary work pieces, if it has less the image noise interference, and filter the detected noise points through improved median filtering method. If the image has larger noise, through the analysis of the process in lowering the quality of the ray images, according to the detector in the case of linear shift-variant system, the image can be optimized by using the algorithm of approximate point spread function and based on the inverse filtering signal to resume, and then the image can be filtered by KNN algorithm. For image of the work pieces in motion, the image offset can be estimated according to the speed of moving work pieces and the size of collected image and so on. Moreover, it can be enhanced by superimposing the offset frames. By means of experiments, the system can not only achieve real-time efficiency but also ensure the quality of the results and realize the dynamic contrast-enhanced of images.