An Image Filtering Method Based on Fractional Differential Estimation Gradient Domain

Abstract

The invention discloses an image filtering method based on a fractional order differential estimation gradient domain. According to the method, in the value constraint Ed aspect, median values in a mask area are utilized to carry out value constraint; in the gradient constraint Eg aspect, gradient constraint based on integer order differential and edge constraint based on integer order differential are fused to obtain a gradient based on fractional order differential; when a direction histogram is calculated, Gauss weight of each direction area is employed to obtain a final gradient direction descriptor, and the direction histogram is constructed for direction constraint. The method takes the fractional order differential as a base, a relatively comprehensive and accurate filtering framework is obtained, so image filtering is better realized, and image filtering quality is improved. The method is applied to image light supplement, image de-noising and image sharpening, and a signal-to-noise ratio, the average gradient and the average information entropy of output images are higher than those of a traditional filtering framework.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to an image filtering processing method, in particular to an image filtering method based on fractional differential estimation gradient domain. Background technique [0002] For a long time, researchers hope to design an image filtering framework with significant filtering effect and stable filtering method, and can break through the traditional image filtering method, and can better enhance the texture details of image edge areas and flat areas while filtering. The image filtering methods are mainly divided into two categories: transform domain and space domain. In space domain, enhancement algorithms based on image gradients are widely used. [0003] Now common image filtering methods are as follows: 1. The image filtering framework proposed by PRAVIN BHAT constrains the distribution of corresponding pixel values ​​and pixel gradient values ​​of the two images by linking the mappi...

AI Technical Summary

Problems solved by technology

However, this method directly uses the current pixel to do the difference calculation when processing the corresponding pixel value, the constraint on the pixel value is not accurate, and the influence of the neighboring pixels on the current pixel is ignored when the gradient distribution is constrained, and the image filtering effect is not good. Not obvious

[0004] 2. Subsequently, many improved algorithms based on this method appeared, while traditional integer-order differential filter operators such as Sobel operator (based on first-order differential) and Gauss-Laplace operator (based on second-order differential) will make image pixel values The texture detail information with little change has a large linear attenuation, so this kind of edge enhancement operator cannot handle the texture detail of the smooth area of ​​the image well.

The proposed filtering framework underestimates the image information, and does not consider comprehensively when constructing the filtering framework, and does not use fractional differentiation to preserve the low-frequency contour information of the image during the calculation process.

Images