Research On Motion Estimation Method Based On Fuzzy Image

With the unceasing improvement of production technology, the demand for the motionparameters measurement precision of high speed, high frequency and tiny objects is higher andhigher, and followed by considering the problems of measuring equipment performance, volume,cost and the complexity of building platforms, which largely limits its application in practice inthe large degree. In view of this, the measurement of motion displacement based on motionblurred image is studied in this paper.Two different methods of motion estimation are proposed, based on the theory that themotion information are extracted from motion blurred images could be used as the description ofthe motion of the object itself.The first method is that using the motion blur information to estimate the motiondisplacement. According to the motion blurred image information, the object of motion-from-bluris represented by extracting channel information from the blurred images. Using SpectralMatting algorithm to extract the channel information of blurred image and establish a motionblur constrain model which is closely related to the well-known optical flow constraint. Toimprove the measured accuracy, this paper adopts filters method to estimate the gradient of themotion blurred image channel information, and this method has carried on the simulationanalysis by using different image. The experimental simulations show that the precision ofmethod that the filters calculation the motion parameters are superior to the traditional first-orderdifferential method.Another method is used to estimate the motion displacement based on alternated exposure.For the motion blurred images, it’s a good way to avoid aliasing by prolong the exposure timewhich can constrain the high frequency of the blurred image. this paper takes the cover and thetime offset into account to set up the constraint model, through the method of total variationminimization to conduct multi-scale decomposition of the input image, in order to clearly reflectthe detail of the image, the bipolar decomposition method is adopted. The experimentalsimulations show that the method is feasible for motion displacement estimation.Combined the above two methods from the image acquisition complexity, the quantities ofthe compute-need images, the running time and the precision etc. Analysis shows that theα-channel of the motion blurred method is prior to the alternated exposure method.