An Image Representation Method Based on Practical Robust PCA

Abstract

The invention discloses an image representation method based on practical robust PCA, which belongs to the field of pattern recognition. The method includes the following steps: reading the image data set, and establishing a sample matrix according to the pixel values; inputting the sample matrix into a pre-built target model, and the target model is joint learning based on robust reconstruction error minimization and robust data difference maximization model, which projects the data into a low-dimensional subspace according to the transformation matrix W, and uses the restoration matrix W to restore the data, and uses the L2,p norm as the distance measure; the target model is solved by an iterative algorithm based on PCA technology, and the obtained Transformation matrix W; complete image reconstruction according to transformation matrix W. The invention establishes the connection between the original space and the converted space features, and uses the L2,p norm distance measure to weaken the sensitivity to outliers, and improves the robustness of PCA well. In addition, the present invention designs a new iterative algorithm to optimize the minimization problem based on the L2,p norm, and the algorithm has better convergence.

Description

technical field [0001] The invention belongs to the field of pattern recognition, and in particular relates to a PCA-based image representation method. Background technique [0002] Various high-dimensional data, such as images and texts, are often encountered in practical situations. How to effectively represent such data has always been one of the most important issues in pattern classification. Feature extraction (or dimensionality reduction), as a useful data analysis tool, has been widely used to address this problem. Principal Component Analysis (PCA) is one of the most representative techniques. PCA performs feature extraction and image reconstruction by finding the optimal projection vector to maximize its variance or minimize reconstruction error. [0003] When the computer reads the image to obtain the data matrix, the existence of outliers or noise is very common due to many reasons (such as lighting, occlusion and other factors in the original image or hardware ...

AI Technical Summary

Problems solved by technology

[0004] For high-dimensional data such as images or text, the model optimization goal of traditional PCA is to maximize the data variance or minimize the reconstruction error. Theoretically, these two forms are equivalent under the square L2 norm distance measure, but The disadvantage is that the feature extraction effect is not good in the face of outliers or noises that are prevalent in the image data set.

Although under the robust norm measure, such as the L1 norm, the robustness of the model in dealing with outliers is improved, but the minimization of the data variance and the reconstruction error cannot guarantee the equivalence. However, the two are effective for feature extraction. all play a vital role

Images