| Images are one of the most important sources of information for people to know and understand the world.The level of image resolution means how much information the image contains.High-resolution images have higher credibility,clearer content,more accurate information expression.High-resolution images are clearer,express information more accurately,and have higher credibility.Three-quarters of the total amount of information humans obtain comes from the visual system.Scientific instruments and equipment often express information by generating images rather than generating sounds or smells.With the development of science and technology,both in daily life and high-tech fields,higher quality images are needed to meet people's needs.However,the image acquisition process is affected by hardware facilities and environmental factors in practical applications.The resulting images often have problems such as distortion and degradation,which lead to unsatisfactory visual effects.Therefore,transforming low-resolution images into high-resolution images through image enlarging technology to meet people's needs is a research hotspot in the field of image processing.High-frequency information areas with strong pixel mutations such as edges and textures contain a lot of effective information in the image.It is also the part of the image that attracts the most attention.How to maintain the integrity and authenticity of high-frequency information is a difficult point in algorithm research in the process of algorithm implementation.Based on the analysis of the above-mentioned problems,this article has conducted the following research:(1)Aiming at the problem of image edge blurring of traditional interpolation method and the small improvement of image quality by iterative back projection method.This paper proposes a rational polynomial image enlarging algorithm with feature constraints.Since the image pixels can be regarded as sampled in the unit area of the original scene surface.The original scene can be approximated by a piecewise quadratic polynomial.For each pixel of the original image and 8 pixels in the neighborhood,a fitting surface patch is constructed with edge features as constraints.Then splice all the surface patches into an interpolated surface with the characteristics of the original image.The sampling formula is redefined by interpolation surface to enlarge the image.The method in this paper adds the consideration of the approximation accuracy of the surface patch,while the previous definition of the weight function only considers the distance factor in the process of surface patch splicing.For the error of the curved surface fitting image block,the high frequency information lost in the interpolation process is compensated by the back projection technique,and the minimum error between the iterative convergence result and the input image is used as the basis of back-projection.The method in this paper has been greatly improved in both subjective vision and objective quantification,reducing the jagged and mosaic effects of the enlarged image edge,effectively improving the image quality,and better retaining the edge details in the image.(2)In order to solve the problem that the edge width expands with the image enlarging during the image interpolation process,the edge is blurred,and the sharpness cannot be effectively maintained.This paper proposes a hybrid polynomial image enlarging algorithm based on Canny operator.The high frequency information of the image is extracted through the improved detection operator,and the set of high-frequency information is added to the interpolation polynomial structure to improve the accuracy of the interpolation polynomial and realize image enlarging.Then,based on the high-frequency information of the image,the edge of the enlarged image is detected and optimized to ensure that the enlarged image has a clear visual edge.The method in this paper improves the edge definition,maintains the smoothness of the flat area,and effectively improves the image quality. |
PDF Full Text Request