Image filtering method and device, terminal and computer readable storage medium

Abstract

The invention belongs to the technical field of image processing, and mainly provides an image filtering method and device, a terminal and a computer readable storage medium. According to the invention, by introducing an error threshold value of each pixel in a to-be-filtered image is introduced and comparing the pixel difference corresponding to each pixel in the to-be-filtered image with the error threshold value of the corresponding pixel in the to-be-filtered image, corresponding pixel values from an initial filtering image and the to-be-filtered image are selected as corresponding target pixel values of the pixels in the to-be-filtered image in a target filtering image, therefore, the problem that flying spots appear in the target filtering image obtained after the to-be-filtered image is filtered is effectively avoided, the target filtering image with a good edge preserving effect and a good filtering effect can be obtained, and compared with a guiding filtering mode, the method has the advantages of being small in calculation amount, low in hardware requirement and capable of being conveniently achieved on various calculation platforms.

Description

technical field [0001] The present application belongs to the technical field of image processing, and in particular relates to an image filtering method, device, terminal and computer-readable storage medium. Background technique [0002] In the image filtering algorithm, guided filtering, bilateral filtering and least squares filtering are called three edge-preserving filters, which belong to anisotropic filters. smooth. Compared with bilateral filtering and least squares filtering, guided filtering can be used to process images with more image noise, and the edge optimization effect is better. [0003] However, guided filtering requires a large amount of calculation. Even if the downsampled image is used for fast guided filtering (FastGuided Filter), the calculation amount is still very large, especially in the calculation process, in addition to a large number of additions and multiplications, division operations are required. , its realization cost is high, it is diff...

AI Technical Summary

Problems solved by technology

[0003] However, guided filtering requires a large amount of calculation. Even if the down-sampled image is used for fast guided filtering (FastGuided Filter), the calculation amount is still large, especially in addition to a large number of additions and multiplications during the calculation process, division operations are required. , its implementation cost is high, it is difficult to implement it with general processor CPU or DSP, even if it can be implemented with ASIC special circuit, its cost is also very high

Especially for floating-point data, if fast-guided filtering is used to calculate the variance and mean value of the down-sampled image to achieve filtering, the filtered image and the value of the initially input floating-point data image are on the edge of the object. Flying spots will appear in the transition areas such as the junction of the foreground and the foreground, which reduces the quality of image smoothing and filtering

Images