Image decomposition method and device

Abstract

The invention discloses an image decomposition method and device, wherein the image decomposition method includes the step of: calculating the direction feature of the target two-dimensional image according to the gradient of the target two-dimensional image in the x-axis direction and the gradient in the y-axis direction quantity and periodic complexity; calculate the structural metric of the target two-dimensional image according to the image gradient of the target two-dimensional image and the directional feature quantity; calculate according to the image gradient of the target two-dimensional image and the periodic complexity A texture metric of the target two-dimensional image; constructing an objective function corresponding to the target two-dimensional image according to the structure metric and texture metric, so as to decompose the target two-dimensional image by optimizing the objective function. The invention can preserve the structural edge of the image very well, and can remove the periodically repeating texture in the image.

Description

technical field [0001] The embodiments of the present application relate to the field of image processing, and in particular, to an image decomposition method and device. Background technique [0002] Image structure texture decomposition refers to dividing the original image into structural components containing main information and texture components containing small and messy details. It is an important image processing technology and is widely used in many fields in image processing. , such as image enhancement, image denoising, image repair, edge detection and image quality evaluation and other fields. [0003] The existing structural texture decomposition methods are mainly divided into three categories: the first category is the decomposition method based on variational optimization, the second category is the filtering-based decomposition method, and the third category is the neural network-based decomposition method, in which the variational The optimized decomposi...

AI Technical Summary

Problems solved by technology

For example, the decomposition method based on variational optimization cannot distinguish fine structures and textures, and the rolling-guided filtering method relies too much on the scale of the image content. In fact, in the same image, different structures can have different scales. Another On the one hand, the unsupervised deep image smoothing method obtains a deep neural network model based on a large number of external sample training, but the model cannot achieve good results on many different types of pictures, and its robustness is poor

[0005] In summary, the structure images obtained by the existing image decomposition methods are often over-smoothed, or cannot remove the texture on the structure images well.

Images