Research On Text Detecting And Removal In Video Image Based On CUDA

Video and image have become the main ways for people to access information orcommunication in daily life. Text detection and removal of the video image is a kind ofvideo inpainting technique, which is a effective recycling of video images resource. Withthe emergence of high-definition video images, the amount of video and image data andthe wealth of information have increased dramatically with the widely application.Research of video image real-time processing method has become the urgent needs ofvideo image application. This thesis focuses on how to improve the processing speed oftext detection and removal for vedio image with GPU.At present, more and more attention is paid to the computing power of GPU. NVIDIA’sCUDA has well programmability and real-time processing capabilities for image. In thispaper, we analysis of the CUDA architecture and parallel processing mechanism bystuding the working principle of CUDA parallel computing.Images text detection andremoval is the classic content in the digital image processing.Text detection has animportant significance for information extraction and search of video image. This paperintegrates the traditional algorithm and put forward a kind of text detection and removalalgorithm for single image. Because the text and content of the video with a certaincontinuity of the video from frame to frame, we put foword a algorithm of vedio imagedetection and repair to improve the video text detection accuracy and image restorationeffect. In this paper, parallel modules can be carried out in CUDA processing, which canreduce the run time of the algorithm.we can also optimize some modules in the programaccording to their data characteristics.The video image text detection and removal system achieved good video imageinpainting results for this paper. Compare the program’s running time of GPU and CPU,comparative tests show that there is better optimization effect by CUDA acceleration.