Image Interpolation Based On Edge Information Method

The digital image interpolation in the field of engineering andmathematics has aroused widespread concern, and the imageinterpolation technology has been widely used in digital image processing,such as image zoom, image deformation, image restoration, imagereconstruction. Image interpolation is known to use the adjacent togenerate unknown gray value of the gray pixels value to the originalimage from renewable to a higher-resolution image.In this paper, firstly, we summarized the classical image interpolationalgorithm and that has been raised in recent years. As the traditionalclassical interpolation algorithm, there is more obvious aliasing andsmoothing effects. For such problems, this paper proposed an imageinterpolation based on edge information methods (L-BEI). The basic ideaof the algorithm is set in advance a suitable threshold value to divide theimage into the non-edge region and the edge region, and fullyconsiderated two adjacent pixels in the edge region of the image points inthe four directions of the pixel value difference (horizontal and verticaldirections, the two diagonal directions), according to the magnitude of thedifference,we can determine the correlation maximum of the correlationof adjacent pixels, the direction which has the smallest difference has thehighest correlation in the four directions. Secondly, we consider the caseof the image interpolation which enlarges more than twice will increasethe error. Meantime we give full consideration to the characteristics of thelocal structure of the image, image interpolation which obtained the newpixel is to be interpolated points, but for the adjacent pixel, its pixel valuemay be two cases. For this problem, the paper proposes twicemorphology operations on the edge of the image, and then using themulti-directional correlation characteristics of the edge pixels, therebyreducing the accumulation of errors. The main advantage of H-BEIalgorithm is that its calculation is relatively simple, and the implementrelatively easy, and the efficiency relatively high. The selected piecesclassic test image which on the basis of the edge information, the imagerespectively interpolated by the nearest neighbor interpolation, the bilinear interpolation, and the cubic convolution interpolation, and theNEDI interpolation, and the proposed algorithm in this paper. And thenwe can make a comparison from the subjective visual effect, the peaksignal-to-noise ratio, and the running time.The results show that, in this algorithm, the aliasing is well improved,while ensuring the higher running time, but also better inhibition of thesmoothing effect of an image.