Why is there a difference in texture quality between the entire model and a small single model

We are planning a large project with taken using nearly 700 drone images. Based on the collected model, the model was produced through the Normal mesh operation after alignment. (120M tris)

After setting the texture target image quality to 16K and the maximum number of textures to 10, the texture step was performed to obtain 5 textures.

However, the resolution of the texture work calculated at once is not so accurate that the resolution of the entire calculated texture work is significantly comparable and the results are bad compared to the results of only one building in the same project.

I understand that you are using the Unwrap function at this time, so please tell me how and why you use the Unwrap function, and why the accuracy of the texture of the result calculated only a part of the project is different than the result calculated at once.

Hi yunsik79,
where are you comparing the texture quality?
Can you try these settings:

1 Like

The image is not attached, so there are data and texture settings that I compared in Google Drive. Please refer to the link.
( Model Comparison Analysis Data : Model Comparison Analysis Data.zip - Google Drive)
Of course, I’ll try the way you answered.

Please let me know what’s wrong with the setting now.

Additionally, what is a gutter?

We are not allowed to download data from external sources. But if you are comparing the textures in the RealityCapture, there is a restriction. 1. Textures showed in the 3d view are downscaled (if RC detects you want to view more than 116k texture or 48k texture) textures will be automatically downscaled for viewing purposes.
In order to view it in full resolution please position the camera in the 3d view and make a render to view it in full quality/ resolution.

Gutter represents a number of pixels of a chart border in a texture image. Gutter is important for avoiding visible artefacts on a model during rendering. Many rendering engines use techniques like mip-mapping, where the texture is downscaled several times to create an image pyramid. If the engine renders texture from higher pyramid levels, then a color from one chart can bleed to another chart in the interpolation process. This can happen just in case you do not have the gutter between the charts large enough. The default value 2 is enough to avoid this type of artifact in most rendering engines.

Thank you.

I didn’t understand it exactly because I didn’t have related knowledge, but I think there’s a way to solve it.

I moved the PC and attached the picture, and it was attached normally.

I would appreciate it if you could refer to the image below and let me know the problem with the current textures.

From top to bottom, it’s a full model, a full model (enlarged), a small single model, a small single model (enlarged), and the current texture setting.

Both comparison images are only displaying the same area as a clipping box.




An Model(Zoom)

As I wrote before, Textures showed in the 3d view are downscaled (if RC detects you want to view more than 116k texture or 4 8k texture) textures will be automatically downscaled for viewing purposes.
In order to view it in full resolution please position the camera in the 3d view and make a render to view it in full quality/ resolution.
Big model has 5 16k textures, small model just 1 16k texture. In 3D view it is textures just as 1 16k texture, so basically, big model is 5 times decimated related to small model’s texture. To see correct texture you need to render the view for the model.

1 Like