I have an image mapped onto my object and it looks good but when I move the camera very close it looks pixelated and blurry. Not sure why. The image was taken from cg textures. Someone on a forum says change the sampling in the material colour to none so I diid this and it makes no difference. I tried several textures of different sizes and they all look the same when up close. How do I make it clear?
You asked “how to make it clear?” With a scene file (pretty much: always), “saved with assets”. I can provide an upload link if the file won’t work as “zip” attachment. (see in your PM “mailbox”) A scene files might answers so many questions…
I assume that we talk here about the rendering. If you have a texture that looks blurry without rendering, just in the preview, and you need a better way of seeing what is going on, go to that material, to “Editor” [left side of the material editor] and then to Texture Preview Size. But do that only if really needed and as low as needed. It eats resources!
The part that you “see” with the camera should be at least 1.5 times higher in resolution (pixel-amount) than the area in which it is rendered. Think of it that way, no pixel of a texture one a 3D object will show up as one and the same pixel in the render, at least the chances are pretty low.
Blurry and pixelated indicates mostly a non sufficient resolution. It is the real number of pixels not anything dpi or such.
Let me know if you can reduce the scene to camera, object and texture, and how you would like to share it.
Yes Im talking about in the render. I already changed the preview size in the material editor. I need to know how do I make my camera very close to the object without the texture becoming blurry in the render. With sampling set to none its pixelated and with sampling set to MIP its blurry. I have tried to upload a zip but seems Im not allowed. Where shall I send it?
Bikhu, I have send you an upload link to your Cineversity PM (private message).
The preview settings have no infleuence on the render results, just to be clear.
If the resolution is not sufficient, then there is no way.
If the camera comes close and/or the field of view becomes smaller, the resolution of the texture must be higher then, if it turns blurry. There is no way around, if we talk about quality. (up-rez. Is not a solution, it needs to be genuine information)
Please upload the file and I will check in the morning (L.A. time here)
I’m glad I asked for it. Scene files can show much. I have checked all, and found the answer in this case has two parts.
The image with the UV grid on it should be 2286x3194 to provide the suggested 1.5 more information than visible in the render. Of course twice or even four times is always better. There might be always a change—latter on.
If you work with really large textures, perhaps in 32bit/float, then a closer look is needed—if the scene renders slow.
The part that is seen like the right upper corner of the cube counts, and I used here a quarter of the cube to measure. You can see that the left side is more in the background, so tiny little bit less in need for a higher resolution.
To use 8bit/channel textures is perhaps in few cases sufficient, but the sRGB color space is not really something I would suggest in 8bit/channel for any professional use. It is a deliver colorspace for the web, and it has only in the middle area some more tonal values. The risk of banding and comes to mind. There is always a confusion between deliver format and production needs. The color space sRGB is relative close to REC 709, but to produce in it is dangerous. Yes, many people love to stay inside that, but soon REC 2020 is standard and things will look old, very old then, if produced in REC709. Again, delivery vs production pipeline, it is a big difference.
I mention that, as 8bit/channel was OK in the ‘90s, but today I would strongly advice not to start a texture library with this “thin” format.
I hope the two images below allow for some clarification. Personally, I use a checker board [image #3], not as small as attached, but it helps even in this size to see the problems quickly.
• Take the texture [template] that you have, the PSD file, change the image size to 2286x3194, or larger.
• Use the ruler in Ps or the measure the size you need to fill the 4x4 UV grid for the cube side in question.
• Compare this size with the texture image you have. If the texture {pattern] is larger, perhaps upscale the Ps file with the ._UV information so you don’t scale the original texture down to make it fit into it.
• Then save that file with a different name and replace it in the material editor.
• You should see a proper result by then—if not please let me know.
.
(One thing that is normally not discussed, but worth mentioning, if this object is seen through a refraction, magnification, or is reflected in a way that parts of the object can be seen bigger, the resolution must follow this then as well. Mostly this is the case with 360º/180º textures for environments and then seen on an concave (bended) reflective object.)
Ok thanks so what Im basically understanding from what you are saying is that I need my texture image to be higher resolution. Not sure why i can’t scale it down though? Usually there is a problem with scaling up but scaling down I didnt think would be an issue.
Do the following, to see the result (not a suggestion for normal work, just a test):
Open an image in Photoshop. Now press “option/alt, command, I” or Image>Image size. Use the Percent. Set the Percent to 10% for width and hight. Click OK. Now you have “scaled” the image down”.
Do it again and set it now to 1000% percent. It brings the image back to its previous resolution. This would be the step the C4D camera does, if getting closer. If you can’t see now the huge degradation of the image, go to the Navigator and set the view to 100% so one pixel in the image is one pixel on the screen. If you like go to the History and click to the top entry (open image), and then to the lowest, Scale. Now the change should be obvious.
If that doesn’t convince you, do the same, open an image, make a copy (Layer) and do the same with the layer, scale to 10% then to 1000%, compare the two.
There is not lossless scale down option in C4D. In Photoshop that would be possible while you use first the option to put it into an container, called “Smart Object”. This can be scalled up and down as much and as often you like.
If saved or merged, that option is gone, and if it was scaled down, the main resolution is gone!
A single pixel is a specific information, any change to it, will alter it and in any way the quality goes down. The only exception is: a pixel moves a full pixel size—one pixel or hundred pixel in x or y. Anything else will damge the quality, scaling of deforming (lens-correction) will never improve the quality, given the initial quality. Scaling up high and sharpening is sometimes suggest in some weird tutorials, then scale it a little bit down. This is worse case scenario, as the sharpening by itself destroys information, and is then useless for rendering.
Again, each pixel should be left alone as much as possible, if we talk about the use in 3D. There is never an improvement possible if mixed or scaled. Never. If scaled down, a huge part of the information is gone, forever.
The UV mesh is based on a normalized space. Which goes ALWAYS from 0 to 1. No way to increase that. Only inside of that space changes will create changes. With that, you can use small “proxy” images and replace them later with higher resolutions. So there is no trick, or as some believe even to set the dpi higher and everything will work, not true.
To scale the UV will do nothing to increase the resolution nor the use of a texture, if the images should look like in the same size.
Pixels are information, and that is the only thing that counts. The more pixels the more information, but only if the pixels are not extrapolated, or scaled up. The bit depth of the pixels and the used color space are important as well, Adobe RGB is certainly much better than sRGB, etc.
The area of the object and the amount of pixels is the first critical step, the camera (as described above) with its rendering settings (see images above) are in team work on that.
You need this information genuinely. There is no trick to increase that.
The UV mesh is based on a normalized space. Which goes ALWAYS from 0 to 1. No way to increase that. Only inside of that space changes will create changes. With that, you can use small “proxy” images and replace them later with higher resolutions. So there is no trick, or as some believe even to set the dpi higher and everything will work, not true.
To scale the UV will do nothing to increase the resolution nor the use of a texture, if the images should look like in the same size.
Pixels are information, and that is the only thing that counts. The more pixels the more information, but only if the pixels are not extrapolated, or scaled up. The bit depth of the pixels and the used color space are important as well, Adobe RGB is certainly much better than sRGB, etc.
The area of the object and the amount of pixels is the first critical step, the camera (as described above) with its rendering settings (see images above) are in team work on that.
You need this information genuinely. There is no trick to increase that.
All the best
I just re read that and I see that you say to upscale the PS file with UV info. So then I add the image after I upscale the PS with UV right? So Im basically aiming to enlargen my PS doc to match the larger image? Does this work in reverse? Can I make my PS with UV info smaller to match a smaller image? Not really sure what that checkerboard is for?
Post #5:
Upscaling was for the template only, not for the image. The image needs to be placed after this in it—but only if you have a higher resolution by then. To paste and scale the image up will just make things worse. Some starters might tell you scale up and sharpen—my opinion about that: run as fast as you can away from those people! :o)
Post #7:
The upscale idea was only to show that it will NOT work. Never has, never will (even some crime TV shows like to show it. It will have not the quality that you like to use in CG work. The only option to “enhance” is given from video footage [or image sequences], where all images are merged and analyzed to create more data. Again not the way at all, if quality is the key) If the information is not given you can’t magically introduce details.
=======
There is no option in any program to produce new information. Upscaling a texture will ALWAYS reduce the quality.
I’m aware of upscaling/up-rezing programs, but the simple truth is, what is not in the low res, will not be in the hi-rez. It is as simple as that.
Again, get the images from the start in the proper resolution. Otherwise why bother and buy a 20 Mp or 50 MP camera, when those things would work?
UV scaling will not help it either. To scale up the UV templade that you have will, but only if the images used are not upscaled to fit into it. UV and UV templade ( the one you have had created) are two different things, with complettely different results while scaling.
The checkerboard is to “see” what resolution you need. each square has the same amount of pixels, so the largest magnification in the rendered image allows then to determine the needed resolution in the source image [source image, the one that goes into the material]. Yes, it might be a complex information flow, but it needs to be understood, to get proper results.
In any case, altering the image in any way will reduce the quality. Please never ever use sharpening for textures, that is worse than to leave it as is.
Sorry. No way—only good quality from the start, and keep that quality. Merging images into an UV map, which requires scaling of any kind [up/down] will not help. If not possible, work with several UV tags and texture tag/material combinations. Allways better to leave images as you get them, if they are not way too large.
