A Method of Identifying Diseases of Ancient Murals Based on Global Dictionary Features

Abstract

The invention discloses a method for identifying diseases of ancient murals based on global dictionary features. The method includes the following steps: using a dictionary training method to use an online method for an original mural image based on a pre-identified true value image from a diseased area and a non-disease area. The dictionary learning method trains the diseased dictionary and the non-diseased dictionary respectively; uses the superpixel method to superpixelate the target image; establishes a Bayesian model to identify each superpixel block. The use of superpixels and sparse coding proposed by the present invention can perform image segmentation more quickly and effectively. In the disease segmentation of ancient mural images, using the same Bayesian model, superpixel-based sparse coding can make the algorithm run 103 times faster, which is close to real-time segmentation.

Description

technical field [0001] The invention relates to the fields of computer image processing and pattern recognition, in particular to a method for identifying diseases of ancient murals based on global dictionary features. Background technique [0002] As one of the earliest forms of painting in human history, mural painting is the witness of human history and civilization and the carrier of ancient and modern cultural inheritance. With the passage of time, large-scale murals have been damaged due to various natural and human factors. Therefore, the protection of murals has always been an area of ​​exploration. Finding and taking targeted measures against different diseases is the key . The traditional manual drawing of diseases is obviously inefficient and difficult to store and update. How to better and more efficiently find the diseases of murals is of great significance to the protection of murals. At the same time, the continuous monitoring of each mural image can obtain ...

AI Technical Summary

Problems solved by technology

Its disadvantage is that it cannot describe the local distribution of colors in the image and the spatial position of each color, that is, it cannot describe a specific object or object in the image.

At this time, existing technologies such as Deformable Parts Model (DPM), Saliency Analysis (Saliency) [6] , Region Merge and other general detection techniques are not suitable for this field

At the same time, in the existing detection-segmentation problem, well-performing detection algorithms such as sparse coding have the problem of extremely slow speed. Therefore, the present invention proposes an improved sparse coding method (Sparse Coding) [3,4] , and applied to the specific problem of ancient mural disease detection and segmentation, greatly improving the detection results

Images