An Image Block Clustering Method Based on Fourier Spectrum Features

Abstract

An image block clustering method based on Fourier spectrum features. First, two-dimensional fast Fourier transform is performed on each image block to obtain the amplitude spectrum corresponding to the Fourier domain; then, the frequency on the amplitude spectrum is set The direction with the greatest energy is the first direction D, and the number of main spectral line directions is determined as the image block structural complexity C, and the frequency components are calculated and the frequency component marks are set; finally, according to the first direction D of each image block, the structure Complexity C and frequency component labels for clustering. The method of the present invention uses the frequency spectrogram to extract the structural complexity, structural directionality and frequency component distribution of image blocks, so as to design a collaborative similarity metric for joint structure, direction and contrast characteristics, so as to realize fast and detailed image block data collection clustering.

Description

Technical field [0001] The invention relates to the field of image processing and pattern recognition, and relates to the construction and learning of an adaptive dictionary in the inverse problem of image processing, and in particular to an image block clustering method based on Fourier domain spectrum structure and direction characteristics. Background technique [0002] In image processing and pattern recognition, the inverse problem of image processing based on image blocks is very common, including image denoising, deblurring, super-resolution reconstruction and image restoration. The solution of this kind of problem usually adopts the learning algorithm to learn the signal of the image block from the training sample data set, or the image block signal is approximated by a set of basis in the dictionary. Therefore, the performance of the training sample data set or dictionary directly determines the reconstruction result. However, the local content of different regions of t...

AI Technical Summary

Problems solved by technology

However, the existing image block clustering methods usually only consider one or two feature items in the spatial domain, and lack a collaborative metric composed of multiple feature information, which greatly limits the refinement of clustering and the accuracy of various types of clustering. Similarity of image blocks

Images