Image Deformation Algorithm Research Based On Forward Mapping Approach

Image deformation is an image processing technology which transforms local area of a source image to target image. It can be widely used in virtual reality, computing animation, medical image processing, video entertainment and other fields.Mapping approach can be divided into forward and inverse mapping, the image deformation based on forward mapping is studied in this paper. Forward mapping has the following problems: Source point mapping to the target image may be out of borders; multiple points of source may be mapped to one point in target region, it's overlap mapping; target point can not be mapped back to source region, it's a mapping hollow.In this paper, the mathematical model of forward mapping was studied, according to the characteristics of mapping out of borders, we defined the forward mapping transformation formula, deducted the parameters for base functions and shift functions, theoretically solved the mapping out of borders problem.This paper presented an algorithm which uses average theory to solve the overlap mapping problem;At the same time this paper used interpolation methods to solve the mapping hollow problem.With using smallest square sum of RGB values difference to get the most appropriate pixel, it saved the most number of image details; finally eliminated the non-phenomenon of the border squeezing.According with the strategies provided above, this paper presented an algorithm on custom source region, which uses a center point and radius to determine the source regions and target regions under interactive mode.From the implementation viewpoint, this paper proposed to use the scan line to simplify the algorithm. Scan lines can only be processed in horizontal manner, with the use of geometric transformations; it can make deformation done in any direction.Finally, the deformation operation can be controlled, the deformation result is smooth.