When there is a 3D model, the picture will be blurry. Can this problem be improved in the next version? This problem is still very serious.
Now the one on the left has been programmed to sharpen, but there is still a big difference in clarity;
When there is a 3D model, the picture will be blurry.
-
-
- Official Post
Hi,
try increasing the image.cubemapsize setting to 4096:
When using 3D-models and rendering them as 3D-models and not as depthmap, then it's necessary to use a fixed-size cube-map (cube-texture). The pano image and tiles will be then rendered (and scaled) into a that cube-map.
And to be able to use mip-mapping (for better image quality) the size would need to be a power-of-2 size (e.g. 1024,2048,4096,8192). But with the size, also the memory need increases exponentially.
Best regards,
Klaus -
For some mobile phones, this method seems to cause crashes.
try increasing the image.cubemapsize setting to 4096:
When using 3D-models and rendering them as 3D-models and not as depthmap, then it's necessary to use a fixed-size cube-map (cube-texture). The pano image and tiles will be then rendered (and scaled) into a that cube-map.
And to be able to use mip-mapping (for better image quality) the size would need to be a power-of-2 size (e.g. 1024,2048,4096,8192). But with the size, also the memory need increases exponentially.
Best regards,
Klaus -
- Official Post
Right, that can be memory/resources related - cubemapsize=4096 needs 4x times the memory of cubemapsize=2048 (384MB vs 96MB). Additionally more processing is involved in multiple steps...
So it's a trade off...
-
Is there a better way to handle clarity in the next version?
-
- Official Post
Not that easily...
As said it's a trade-off between memory usage and support of older/lower devices.
A real full dynamic solution like 'multi-resolution on images', would require a pre-processing of the 3D-model in a kind of 3D-tiles. Then only the 3D-tiles and their textures that are currently visible can be loaded - and the ones that are not visible can be unloaded/removed when there is not enough memory.
That pre-processing can't be done in real-time during viewing, that would be a offline 'pre-processing' step. I have already thought about that and maybe it will come in future versions as feature of the krpano tools.
But there might be also a little short-time solution: in the next krpano release WebGL2 will be enabled by default (when available) - and with WebGL2 it is possible to use non-power-of-2 textures - and that would allow using cubemapsize settings between 2048 and 4096. So an idea might be to choose automatically a cubemapsize setting that better fits the current pano-image, but is still below 4096...OR - an other solution - use rendermode="depthmap" - with that the 3D-model will get converted to a depthmap and then multiresolution and so the full resolution of the pano-image can be used. But in this case the 3D-model will be less precise.
Best regards,
Klaus -
Or can the program set the user to zoom in on a normal panorama instead of a 3D panorama when zooming the fov to a certain range? Because generally from a normal viewing angle, the user will not perceive blur, but if the viewing angle is enlarged, the blur will be felt;
-
- Official Post
Sorry, I don't understand what you mean... but you could anytime disable the usage of the 3D-model by changing the image.depthmap.enabled setting...
-
Will this clarity problem be solved in the next release?
Participate now!
Don’t have an account yet? Register yourself now and be a part of our community!