Study On The Improved Elimination Algorithm Of Remote Sensing Image Seam Based On Gradient Field

Image mosaic technology have a wide range of applications in Land, sea and air fieldand other fields. Remote sensing image mosaic is to stitch together two or more images,coming from different conditions, eliminating redundant information, constitute a completeimage needed for the operator. In order to obtain a wider range of ground image, needingto splice many scenes into an image, in the remote sensing image processing, due to thegray discrepancy, there inevitably exists a seam-line on the splicing boundary. It is not onlyunable to present complete and actual picture effectiveness, but the nature of the terrainmay be changed, thus affecting the late-stage study. Therefore, how to eliminate theseam-line caused by gray level or color difference after the image is stitched from thegeometry desiderates. At present the algorithm and evaluation of the seam elimination ofremote sensing image are not perfect, and the evaluation problem of the effects of splicinghasn’t been solved excellently. According to remote sensing image and flat-fell seam toeliminate problem, this paper mainly studies the following contents:At first, this paper sums up and summarizes the existing image spell seamelimination methods and expounds their advantages and disadvantages; Introduced therelevant knowledge, the research status quo of gradient algorithm and its backgroundsignificance, detailedly explained the principle and the solving process of the gradientalgorithm.Secondly introduced the pretreatment of obtained TM image, and gives the flow chart;and data related to the band’s characteristics; given two algorithms given are least squaresapproximation method and two domain method on how to determine the image overlapareas of the image and given their application conditions. The above laid a foundation forgradient algorithm to the realization.At last, on the basis of the existing image gradient field to eliminate seam-line, usedthe second-order difference scheme to discrete the energy function of gradient algorithm, atthe same time compared the images gained by gradient algorithm to the images gained bythe algorithms of iteration, and the hard correction method for splicing images, moreovercomparison and research the images obtained by gradient algorithm dispersed for first-order and second-order difference scheme, visually can be seen the image effect dealtwith gradient algorithm processing is better. In addition to meet both the human visualcharacteristics and real gray value features, to satisfy the demand of user for imagematching the visual evaluation and objective evaluation, this paper gives the imageevaluation of subjective and objective evaluation method and the subjective and objectiveevaluation method of splicing image, and further improved the objective evaluationmethod of using the variation coefficient calculation method for testing the effectiveness ofthe proposed criterion, combined with the image gray level histogram to emerge the imageeffect, making the image evaluation standards more effective.