| Summary: | 碩士 === 國立中興大學 === 機械工程學系所 === 105 === Digital image processing measurement is a non-contact and non-destructive optical measuring method.
In this paper, capture the image of object attached to the grid after deformation, and then the object will be restored to the undeformed shape through a series of image processing .The first is to using the blobs detection to find the grid of highest response as the initial restoration point, and using the Canny edge detection to decrease the amount of data. For the initial restoration point, The affine transformation matrix obtained by affine transformation is used as the initial deformation amount. In other grid points, using the concept of image morph to calculate the image of the restoration direction by randomly giving each block a restoration direction, and compare with the standard image to calculate the image structure similarity (SSIM). The restoration amount of the highest similarity is regarded as the restoration direction of the image,and then gradually iterate along the gradient of direction to obtain the highest deformation of the image similarity.
The image similarity used in this paper ignores the similarity of illumination, and only calculates the contrast similarity and the structural similarity to eliminate the factors that image illumination affects the result when the image is captured. Each deformation choosing will use remapping to produce the image of the block to compare the similarity, After obtaining the highest deformation, the blocks generated by this deformation are attached to the corresponding positions.When each block is calculated, it will get an image before the object is deformed.
Keywords：Morph、Affine Transformation、Remap、Blobs Detection、SSIM
|
|---|
