Method and system for generating 2.5 D model by using single picture

Abstract

The invention discloses a method and a system for generating a 2.5 D model by using a single picture, and belongs to the technical field of computer image processing. The method comprises the following steps of: S1, converting an original picture into a depth map by using AI Studio software, then storing the original picture and the depth map together in an asset format of Unreal Engine (UE), and creating a material ball of the original picture on the basis of the asset format; and S2, constructing a dynamic model in the UE, and adjusting the deformation degree of the model according to information provided by the depth map. S3, the deformation strength of the dynamic model is adjusted in real time to optimize the stretching effect of the dynamic model, then the original picture is attached to the dynamic model, and finally the smooth dynamic model is obtained. And S4, creating a static model, storing the static model as an asset file, and then converting the previously made dynamic model into the static model, thereby completing the generation process of the 2.5 D model. The method is characterized in that the scene can be quickly constructed, a large amount of data annotation and training work are not needed, and the method does not depend on professional hardware facilities.

Description

Technical Field

[0001] The present invention relates to the field of computer image processing, and in particular to a method and system for generating a 2.5D model by using a single picture. Background Art

[0002] 2.5D model technology is a special form of visual expression, a model representation method between 2D and 3D. In the fields of computer graphics and computer vision, 2.5D models are often used to describe planes or surfaces with partial three-dimensional effects. Unlike traditional 2D models, 2.5D models can contain some three-dimensional information, such as height, depth, or concavity. These models are usually composed of two-dimensional planes or surfaces, but their surfaces can have a certain degree of three-dimensional effect visually. Compared with traditional 3D models, 2.5D models have higher visualization efficiency and lower computing resource requirements.

[0003] Currently, there are several ways to make 2.5D models:

[0004] 1. Using drawing software to create 2.5D effects has the disadvantage of huge time and labor costs, and the effect of the model produced is limited by the level of the producer.

[0005] 2. Generate 2.5D models using pictures. There are several specific ways to achieve this:

[0006] 1) Methods based on deep learning: The advantage is that it can automatically learn and extract features from images to generate a more accurate 2.5D model. The disadvantage is that a large amount of training data needs to be annotated and the computing resources are high;

[0007] 2) Methods based on optical principles: The advantage is that the depth information of the image can be directly calculated through optical principles to generate a more realistic 2.5D model. The disadvantage is that it requires professional equipment support and has certain requirements for ambient light;

[0008] 3) Pixel-based algorithm: The advantage is that the algorithm is simple and does not require a large amount of labeled data. The disadvantage is that the generated 2.5D model may not be accurate and smooth enough, and is easily affected by noise and image quality. Summary of the invention

[0009] The object of the present invention is to overcome one or more deficiencies of the prior art and to provide a method and system for generating a 2.5D model using a single image.

[0010] The objective of the present invention is achieved through the following technical solutions:

[0011] A method for generating a 2.5D model using a single image, the method comprising the following steps:

[0012] S1. Generate a depth map from the original image using AI Studio software, then save the original image and depth map as UE (Unreal Engine) assets, and create a material ball of the original image using these UE assets;

[0013] S2. Create dynamic model specifications: set the size, set the number of subdivisions of the model, set the deformation strength of the model, and displace the dynamic model according to the previously generated depth map;

[0014] S3. By setting the deformation strength of the dynamic model and configuring the material ball of the original image into the dynamic model, a smoothed dynamic model is obtained;

[0015] S4. After creating the static model and saving it as an asset, synchronize the data of the static model and the dynamic model, bake the mesh data of the dynamic model to the static model, and finally output the static model as a 2.5D model.

[0016] Furthermore, the grayscale value of each pixel in the depth map is set to the depth value of the corresponding point in the scene.

[0017] Furthermore, the dynamic model specifications include: height, depth and concavity.

[0018] Furthermore, the material ball includes the color, texture and glossiness of the image.

[0019] A system for generating a 2.5D model using a single image. The system is used to generate a 2.5D model by inputting a single image.

[0020] The beneficial effects of the present invention are:

[0021] (1) A relatively accurate 2.5D model can be quickly generated using a single image, and scenes can be quickly built without requiring extensive annotation and training;

[0022] (2) No professional equipment is required as support, and the generated 2.5D model is smoothed and not easily affected by noise and image quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flow chart of a method for generating a 2.5D model using a single image;

[0024] Figure 2 is the original picture;

[0025] Figure 3 It is the depth map after processing the original image;

[0026] Figure 4 It is the processed 2.5D model. DETAILED DESCRIPTION

[0027] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0028] See also Figure 1 , provides a method for generating a 2.5D model using a single image, the method steps comprising:

[0029] A method for generating a 2.5D model using a single image, the method comprising the following steps:

[0030] S1. Generate a depth map from the original image using AI Studio software, then save the original image and depth map as UE (Unreal Engine) assets, and create a material ball of the original image using these UE assets;

[0031] AI Studio is a powerful image processing software that supports layer segmentation and depth map processing. It can generate unique 2.5D image effects through layer segmentation and depth effect, which is suitable for professional design work. If you want to generate a depth map: 1. Import the original image into the AI ​​Studio software; 2. Click the Depth Map button; 3. Save the generated depth map.

[0032] Material ball: Often used in modeling software to define the appearance of an object, including properties such as color, texture, and glossiness.

[0033] S2. Create dynamic model specifications: set the size, set the number of subdivisions of the model, set the deformation strength of the model, and displace the dynamic model according to the previously generated depth map;

[0034] Dynamic model: A dynamic model is a model that can change over time. It can move, rotate, scale, etc.

[0035] Subdivision mesh: Subdivision mesh is a modeling technique that allows you to create more detailed and complex models by subdividing the mesh.

[0036] S3. By setting the deformation strength of the dynamic model and configuring the material ball of the original image into the dynamic model, a smoothed dynamic model is obtained;

[0037] Deformation Strength: In 3D modeling, deformation strength is the strength or size of a deformation effect when it is applied.

[0038] Smoothing: This is an optimization technique that makes the model surface look smoother and more continuous by blurring edges and planes;

[0039] Static model: A static model is a model that does not change over time. It cannot make any movements but can move and rotate freely in space.

[0040] S4. After creating the static model and saving it as an asset, synchronize the data of the static model and the dynamic model, bake the mesh data of the dynamic model to the static model, and finally output the static model as a 2.5D model.

[0041] The following is the process of converting an original image into a 2.5D model:

[0042] See also Figure 2 , upload an original image, process it through AI Studio and generate the corresponding depth map, the resolution of the image is 4069*2048. AI Studio calculates the depth information of each pixel of the original image, and then integrates this information into a depth map.

[0043] See also Figure 3 , the generated depth map is a gray image. Put the original image and the depth map into UE, and UE converts the original image and the depth map into UE assets. Create a dynamic model in UE, set the size, add a subdivision network, replace the dynamic model with the depth map, generate a dynamic model with a concave and convex texture, and set the deformation strength of the dynamic model in real time.

[0044] See also Figure 4 , by setting the deformation strength of the dynamic model and configuring the material ball of the original image to the dynamic model, a smoothed dynamic model can be obtained; after creating a static model and saving it as an asset, the data of the static model and the dynamic model are synchronized, the mesh data of the dynamic model is baked into the static model, and finally the static model is output as a 2.5D model.

[0045] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not deviate from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A method for generating a 2.5D model using a single image, the method comprising: S1. Use AI Studio software to generate a depth map from the original image. Then save the original image and depth map as UE assets, and use these UE assets to create a material ball for the original image; S2. Create dynamic model specifications: set the size, set the number of subdivision grids of the model, set the deformation strength of the model, and displace the dynamic model according to the previously generated depth map; S3. By setting the deformation strength of the dynamic model and configuring the material ball of the original image into the dynamic model, a smoothed dynamic model is obtained; S4. After creating the static model and saving it as an asset, synchronize the data of the static model and the dynamic model, bake the dynamic model mesh data to the static model, and finally output the static model as a 2.5D model.

2. The method for generating a 2.5D model using a single image according to claim 1, characterized in that: The grayscale value of each pixel in the depth map is set to the depth value of the corresponding point in the scene.

3. The method for generating a 2.5D model using a single image according to claim 1, characterized in that: The dynamic model specifications include: height, depth and concavity.

4. The method for generating a 2.5D model using a single image according to claim 1, characterized in that: The material ball includes the color, texture and glossiness of the image.

5. A system for generating a 2.5D model using a single image, characterized in that: Used to execute the methods described in claims 1 to 4, and used to input a single image to generate a 2.5D model.

Images