Transform invariant low rank texture feature extraction and restoration algorithms for architectural decoration surface patterns

Recently, image feature extraction has been an essential topic in computer research. In response to the problem that the accuracy and efficiency of extracting image texture features are still insufficient to meet the practical requirements in applications, this study presents a new transformation invariant low rank texture feature extraction and restoration algorithm. Firstly, the basic contents of image texture features and sparse representation are introduced. Then a new transformation invariant low rank texture feature extraction and restoration algorithm is proposed in view of this. From the results, the research algorithm had a higher peak signal-to-noise ratio of 36.02 dB. The high fidelity criterion value of the research algorithm was 7.04. The structural similarity index of the research algorithm was relatively high, with a value of 0.9146. The average relative error of the research algorithm is 2.327%, the mean square error is 1.327%, the mean absolute error is 7.265%, the root mean square deviation was 0.1123, and the coefficient of determination was 0.9998. The experimental results show that the proposed algorithm has good performance in extracting image texture features and has certain application value in pattern extraction of architectural decoration surfaces. Research can provide theoretical basis and data support for image feature extraction, which is not only of great significance in improving the realism and aesthetics of architectural decoration, but also has a broad application prospect in the field of ancient building restoration, which helps to protect and inherit cultural heritage.