Method for constructing sparse representation redundant dictionary based on ridgelet frame

Abstract

The invention discloses a method for constructing a sparse representation redundant dictionary based on a ridgelet frame, which mainly overcomes the defects of huge redundant dictionary scale, huge calculation amount, low speed and sparse representation incapability of images with line singularity in the prior art. The method comprises the following construction processes of: (1) selecting proper functions to be used as generating functions of ridgelet base functions, and determining the ridgelet frame for constructing generating atoms according to the definition of the ridgelet; (2) carrying out range restriction on each variable in the ridgelet frame; (3) taking proper discrete intervals on each variable according to the restriction range for carrying out discretization to generate the redundant dictionary; (4) adopting an energy threshold value screening method for further reducing and optimizing the redundant dictionary; and (5) blocking images, and carrying out sparse representation on each sub block image by using the orthogonal matching pursuit (OMP) algorithm. Compared with the prior art, the invention has the advantages of good sparse representation visual effect, high calculation speed and high precision, and can be used for image processing and sparse representation in the computer visual field.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a method for constructing a redundant dictionary and performing sparse representation on images with the dictionary, which can be used for sparse representation in image processing and computer vision. Background technique [0002] Sparse representation is one of the key technologies in the field of image processing and computer vision. It refers to the absolute value of the coefficients of most of the basis functions in the linear expansion of the image are close to zero, and only a few basis functions have large non-zero coefficients. And the limited large coefficients can represent most of the information of the image. [0003] In the field of image sparse representation, researchers have proposed many sparse representation methods, including various orthogonal transformations, such as wavelet transform, cosine transform, Fourier transform, etc. Later, researchers put t...

AI Technical Summary

Problems solved by technology

[0007] (1) The size of the redundant dictionary is large, resulting in a sparse image, which means that the search space for effective atoms is large, and the search time complexity increases accordingly, resulting in a huge amount of calculation

[0008] (2) The existing redundant dictionaries cannot effectively sparsely represent rich geometric structures of edge contours, such as straight lines and curves, and the represented images have block effects

Images