Automatic image repair technique of Drawer Box With Trapezoidal Lining and its application

Automatic image repair technique of Drawer Box With Trapezoidal Lining and its application

Images are the main source of information that humans get from the outside world. Since there have been images, there has also been the technology to repair broken images. Inpainting image restoration technique is an ancient art. In the European Renaissance, in order to restore the lost or damaged parts of the works of art while maintaining the overall effect of the works, people began to repair the works of art, which was mainly to fill the cracks or gashes on the works, such work is called Inpainting.

Nowadays, image Inpainting repair technology has become a new active research direction in the field of image engineering. Its purpose is to study and solve how to better realize the detection of the damaged part of the image, and automatically restore the damaged part with the image repair algorithm according to the effective information around the damaged image. Although some powerful digital image processing software, such as Photoshop, can also carry out professional special effects processing and repair processing of damaged images, but it requires experienced technicians to carry out complex manual processing, unable to achieve the purpose of "computer intelligence" automatic processing.

Let's first look at an example of image repair. Figure 1(b) shows a letter symbol like "@", part of which is blocked by an obstacle. According to people's visual experience, and referring to the shape of the obscured object and the surrounding image information, a guess judgment is made that after removing the obstacle, characters as shown in Figure 1(a) or as shown in Figure 1(c) may appear, and the repair results of Figure 1(a) and Figure 1(c) may be correct. Computer image Inpainting technology simulates this effect from the perspective of human visual psychology. According to the edge information of the object to be covered, it extends and diffuses, connects the boundary in a certain direction, fills the covered part, and achieves visual connectivity. Of course, we can also see from the figure that the size of the damaged information (the size of the obstacle) of the image also directly affects people's visual connectivity judgment. From a mathematical point of view, the phenomenon that the image repair results are not unique in the case of insufficient information around the damaged image is a pathological problem, so the image Inpainting technique also needs to be based on the computer vision theory. This problem is solved by a limited design repair algorithm with certain assumptions.

According to the Bayesian structural system, the perfect ideal image obtained from incomplete and deformed data is based on the "best hypothesis" or "best guess" made by the simulated human eye restoration image. This "best hypothesis" is based on the following two important factors:

1. Image data model: How we can get more information from existing image data in the original image.

2. Image premodel: What kind of image model the intact original image should be. For example, if we are restoring a picture of a banana or an apple in a fruit bowl, we will have a preconception that it should have a smooth shape, full of yellow and red.

Therefore, "optimal guess" is to maximize the probability of the latter in the Bayesian probability model, fill in the missing or damaged part of the image according to certain algorithm rules, so that the repaired image is close to or reach the visual effect of the original image.

For the image Inpainting restoration technique of texture structure, such as wood grain, rock pattern, etc., the core idea is to simulate and generate local texture for filling. Because texture is a reflection of the distribution or characteristics of material components, its local shape information can express the commonality of the same texture. For a texture map, any two small pieces of texture are similar, so the method of texture synthesis can be used to repair the image by: (1) Histogram statistics of the color information of each pixel within a set range; (2) Create a weight table to calculate the determination degree of related pixel color information; (3) Evaluate and select the most matched pixels for image restoration; (4) Cycle until the damaged image is filled. In various algorithms for texture synthesis, we should extend the research scope from general texture to directional texture, and further extend it to surface texture synthesis, video texture synthesis and so on.

For the Inpainting technique of non-textured images, currently researchers mostly use the repair algorithm based on high-order partial differential equation (PDE) model. The main idea is to use the edge information of the area to be repaired to determine the diffusion information and diffusion direction, from the anisotropy of the regional boundary to the boundary diffusion. The algorithm can fill multiple regions with different structures and backgrounds at the same time, and there is no restriction on the topological relationship of the patched region. In addition, Manou et al. proposed a rapid image Inpainting technique based on the above ideas. By determining the isointensity line direction at the boundary of the area to be repaired and connecting corresponding isointensity lines with straight lines, the boundary fills the area to be repaired by spreading the neighborhood information of the area to be repaired within the range of the isointensity line. This method has a good effect on simple structure image repair, and the repair time is greatly shortened.

Image Inpainting repair technology has a wide range of applications, the first is for static image crack repair and obstacle removal. In the printing industry, different types of digital images that need to be processed before printing, such as damaged old photographs, scratched transmission negative, or removing obstacles from the image that are not related to the content of the image, can be easily repaired by Inpainting technology. Users only need to simply select the repair scope, the computer will automatically complete the scratch, blank area filling, or remove obstacles after filling the background pattern, can greatly reduce the time and manual workload of pre-press image processing, but also according to the repair results for further image processing.

The second is motion image repair. Inpainting restoration technology can be applied to the film and television industry. For example, for film and television copy, if one of the continuous frames has scratches, stains and other conditions that need to be repaired, we convert the continuous frames of the film into digital image sequence, and then extract useful information from the adjacent frames to copy and repair according to the comparison of the information of the frames before and after. In addition, subtitles in different languages in the film can also be removed by Inpainting restoration technology.

Thirdly, in some photography fields, the red eye phenomenon in digital photos can be eliminated by Inpainting technology, or the overlapping parts in adjacent photos can be repaired by Inpainting technology when the digital photos taken in several times are stitched together into panoramic images.

Inpainting technique has attracted the attention of many scholars in China and has become an active research field abroad. All in all, the technical advantages of Inpainting will certainly be involved in more image applications, such as biomedicine, remote sensing, and promote the development of this technology in different fields.