Image noise reduction method based on Primal Sketch correction and matrix filling

Abstract

The invention relates to an image noise reduction method based on Primal Sketch correction and matrix filling. The image noise reduction method comprises, at least, step 101, adopting a Primal Sketch model to obtain a sketch map P of a noise adding image I1; step 102, providing a two-stage sketch line correction rule and a corrected sketch map P and designing a direction window for a pixel point on each sketch line segment in the corrected sketch map P in the direction of the sketch line segment to obtain a region map, mapping the image I1 into a structural region E1 and a non-structural region E2 according to the region map; step 103, simultaneously utilizing an image block variance statistic method to divide the image I1 into a smooth region E3 and a non-smooth region E4; step 104, respectively marking the smooth region and the non-smooth region overlapped with the non-structural region as a smooth region E23 and a non-smooth region E24; step 105, adopting the image noise reduction method based on the matrix filling to perform noise reduction on the E1 and the E24 and adopting a non-local average method to perform noise reduction on the E23; step 106, merging noise reduction results of the E1, the E23 and the E24 to obtain a final noise reduction result map of the image I1.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to an image noise reduction method based on Primal Sketch correction and matrix filling, relates to a Primal Sketch model, a matrix filling method and a natural image noise reduction method, and can be applied to natural image noise reduction processing. Background technique [0002] There are many kinds of existing image denoising algorithms, although they can achieve the effect of denoising, but almost every method has its own advantages and disadvantages, for example, the non-local mean denoising method has more advantages than traditional image denoising methods. The noise reduction effect is good, but its algorithm complexity is large, and many dissimilar pixels participating in the weighted average will introduce a non-negligible cumulative error; while the three-dimensional block matching noise reduction method (BM3D) can better maintain the similarity of the image it...

AI Technical Summary

Problems solved by technology

But at the same time, BM3D has disadvantages such as cumbersome steps, large amount of calculation and long operation time.

Images