Weld image processing and GPU realization based on mutual-onformation
Shilin Cui · Welding Technology · 2010
In the thesis,it was proposed a improved matching algorithm based on mutual-information,which expanded grayscale image mutual-information matching algorithm to the field of binary image matching while retaining the information amount of the original grayscale image. Its execution speed didn’t relate with the size of the source graphic or template,but only form linear relation with the binarized source image and the sum of front ground spots of the template. The calculated results by GPU showed that the location result of this algorithm was high in reliability and quick in speed,the problems as in the images,like occlusion,noise,deformation,rotation,zoom and illumination change in the image could be effectively overcome,meeting the requirements for real-time seam tracking application.