Image classification method based on probability density distribution dictionary and Markov transfer features

Abstract

The invention discloses an image classification method based on probability distribution parameter features and Markov transfer features. The method includes: firstly, taking Cauchy distribution, Laplace distribution, generalized Gaussian distribution and alpha-steady-state distribution as atoms, building a distribution dictionary, and fitting high-frequency wavelet transform coefficient distribution of an input image under different scales and different sub-bands; respectively calculating Markov transition probabilities of discrete cosine transform coefficients along the horizontal directionand the vertical direction by utilizing a first-order difference operator; taking the probability density distribution parameter characteristics and the Markov transition probability parameter characteristics as classification characteristics, using a support vector machine LIBSVM as a classifier, taking a radial basis function as a kernel function, obtaining the classifier based on the support vector machine LIBSVM and used for screen content images and natural images through training, and then further realizing automatic classification of the screen content image blocks and the natural imageblocks.

Description

technical field [0001] The invention relates to the field of image and video processing of screen content, in particular to an image classification method based on probability density distribution dictionary and Markov transfer feature, which is stable, efficient, strong in adaptability, and high in classification accuracy. Background technique [0002] Moderately complex, high-efficiency coding for screen content video is one of the latest challenging research topics in the field of video coding, and scholars call it "screen content video coding". At present, the compression efficiency of H.264 / AVC and HEVC for discontinuous tone content such as lines, text, and graphics in screen content has not yet met the application requirements, and the calculation complexity is high. The reason is that the screen content is composed of discontinuous tone area and continuous tone area, but the imaging mechanisms and acquisition methods of the two are completely different, making their ...

AI Technical Summary

Problems solved by technology

However, the computational complexity of this method is high, and it relies on the image-specific PRNU pattern noise, and its robustness is not satisfactory

[0007] Overall, the above scheme still has the disadvantages of large classification feature dimension, high computational complexity, poor robustness, and weak adaptability

Also, since the statistical properties of screen videos are different from those of computer-generated images, these schemes are not yet well suited for automatic classification of screen content image patches from natural image patches

Images