An Image Interpolation Method Based On Ramp Edge Model

Image interpolation is an important technology in digital image processing. Traditional interpolation algorithm is simple and easily realizable.They had the same basic principal,that's to find an input pixel,weight another pixel nearby,then output the gray-value,takeing no account of the others.The difference is the way of getting the pixels around.Traditional interpolation algorithm obtains good visional effect,but they can not process the pixel's local character of tempestuousness jump,such as edges, textures,they can lead to the edge detail to be fuzzy in the enlargement image.When people view a reconstructed image,the edge quality takes a significant role in subjective evaluation,so keeping the edges including geometry regularity effectively is very important.More and more researchers have proposed new interpolation methods to preserve image's edges.And this paper classifies the popular methods,the main ideas are as follows:edge preserving,vector quantization,wavelet transform,interpolation kernel,Fractal technology and edge model.Through summarizing and analyzing various methods' characters and shortages,this paper proposes an image interpolation method based on ramp edge model.When interpolating a gray image,the original image's pixels are classified as smooth areas,strong edges and weak edges by edge detecting operator-canny.The pixels of interpolated image are mapped to the original image by some relation,and we judge which type it belongs to.To smooth areas,strong edges and weak edges,we use bilinear interpolation,interpolation method based on ramp edge model and interpolation method based on edge preserving,respectively.Then we reconstruct the enlarged image. When interpolating a color image,the key point is to choose a good color space,and we select CIELa~*b~*,which is vision-oriented and equipment-independent.The experiments show that our algorithm can not only make the enlarged image's edges more natural and clearer,but also obtain better texture effects,and outperform traditional methods,NEDI and Jia-Guu Leu's method in term of average grad.