How to avoid blurry effect? - opengl

I'm currently working on a 3d project with Opengl but i have a blurry effect problem with distant block.
Like you can see, the nearby blocks are fine but the the blocks become blurry very quickly with the distance. (ps: ignore red block)
I tried differents resolutions of image (1024x1024 and 2048x2048) but with same result.
I tried to modify GL_TEXTURE_2D options but not fixed my problem.
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
if someone have an idea how to deal with this problem, it will be good. thx in advance

Posted as comment, now as proper answer so it can get marked as answered.
When looking at textured surfaces, especially distant ones, the frequency of texture features may exceed the necessary sampling frequency (pixels of the framebuffer/screen; see Nyquist limit). To combat this, MIP-mapping is employed, which reduces the feature frequency (minification). However, this does not account for perspective distortion. Anisotropic filtering places samples on the texture in a trapeze to correct for angled textures.
In OpenGL, you need either version 4.6 or EXT_texture_filter_anisotropic. You can then enable anisoptropic filtering per texture with glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT, num_samples) - this sets the maximum number of samples the hardware may use, any given texture fetch may use less though if it deems more to be excessive. Be aware that the upper limit you can set is implementation-defined and must be queried with glGetFloatv(GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT, &max_samples). More samples mean higher quality (reasonably up until 16 or so), but also higher performance cost (not usually a deal-breaker).

Related

OpenGL Super Resolution issues

I have a OpenGL based gui. I use super resolution to be able to handle various scales. Instead of scaling images up, they are downscaled(unless it so happens someone is running at 4000x4000+ resolution).
The problem is, OpenGL doesn't seem to downscale smoothly. I have artifacts as if the scaling is nearest neighbor. (e.g. the text edges are blocky, even though they are not in the original)
These are the settings I use:
glBindTexture(GL_TEXTURE_2D, tex);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
glTexEnvf(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP);
Here is a sample of the artifacts, the scaling is 2:1 I believe. Maybe it isn't exact though due to window edges and such.
You can see the left edge looks perfect(it's not though) but the right edge has weird breaks in it. The original graphic is perfectly symmetrical and no artifacts.
I've tried GL_NEAREST, GL_LINEAR. No mipmapping so...
Surely OpenGL is not that poor at scaling? I'd like something like bi-cubic scaling or something that will produce good results.
I am using OpenGL 1.1. I could potentially pre-scale images but I'd have to do that every time the window sizes changes and might be slow in cpu.
I have jagged edges on some images too. The whole point of super resolution was to avoid all this ;/
Is there some settings I'm missing?
First you have to understand signal theory, namely the Nyquist Theorem (that wikipedia page is overly specific when talking signals in the "time" domain; the principles are universal for all kinds of discretely sampled signals, including images). When downsampling you always must apply a lowpass anti aliasing filter that cuts off all frequency components above half the sampling frequency to avoid the creation of aliasing artifacts. Without filtering even a linear integrating downsampler will create artifacts. The realtime graphics way of implementing a low pass filter for textures are mipmaps. Every mipmap level cuts off at exactly half the frequency of the next higher level.
You have two options now:
Implement mipmapping
Implement a downsampling fragment shader
Of course the sane thing to do would be not to render in an excess resolution in the first place, but render your GUIs at exactly the target resolution.
With the code you provided, i will make guess at what might be the problem.
Try to load your image or at least allocate the memory before you set those texture parameters with glParameteri. And also, set GL_TEXTURE_MIN_FILTER to GL_LINEAR
Perhaps you meant super sampling (SSAA) which use 2 or more times the original resolution and downsample it to get a smooth image?
It does look from your image that it is using Linear filtering (bilinear)
Try using Anisotropic filtering:
glGetFloatv(GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT, &aniso);
glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT, aniso);
Anisotropic filtering can be applied at different levels, this code will apply it at the maximum level, you can use a number less than aniso if you like. These are extention macros, if you don't have the extention defenitions, they are this:
#define GL_TEXTURE_MAX_ANISOTROPY_EXT 0x84FE
#define GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT 0x84FF

OpenGL - Texture Artifacts at a distance despite having mipmaps

Lately, while getting used to C++ (already knowing OpenGL * fairly * well), I've gotten tired of the visual artifacts I see with textures at a distance, especially with large flat surfaces such as terrain. When coding in Java and OpenGL, I saw the same issues as well. Here's what I'm talking about:
I load my textures using the GL_COMPRESSED_RGBA_S3TC_DXT1_EXT (with DXT3 and DXT5 as well) extensions in OpenGL loading, obviously, .DDS images. Each image has 11 mipmaps (1 full-res texture and 11 lower-res textures), each of which are generated with bilinear in Paint.NET. The grass texture in this (where you see most of the artifacts) is 2048x2048. I AM using OpenGL 3.3, with shaders powering almost everything, so I can edit stuff in GLSL.
Image Parameters are as follows
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 11);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_ANISOTROPY_EXT, 16);
In the GLSL Shader, I call texture2D(grassTexture, terrainTexcoord * 512.0f); The '* 512.0f' is so that the grass isn't stretched all over the entire chunk (I'm using chunks w/ heightmap). The sampler is just a generic sampler2D.
And if anyone's wondering, I got the code for the DDS loader from http://www.opengl-tutorial.org/ (best site ever for OpenGL 3.3+ BTW)
I've probably given more can than what's needed but I hate hearing the "Can I see some more code?" Here's what I'd like to know, what is this effect called? What is a good way to fix it? Is there an post processing effect that takes care of this? Am I doing something wrong? (probably the latter, but oh well xD) Thanks!

GL_NEAREST in GLSL?

If I use the fixed pipeline, I can use
glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
to make an image 'pixelated' as opposed to fragments in between pixels in the image being interpolated. How would I do the same thing in GLSL program? I'm using the texture2D function. I ask because I am using a shader program for my skybox, and you can see the edges because the edge pixels get blurred with grey. This problem gets fixed if I were to use the fixed pipeline and the above function calls.
You can use the same texture minification and magnification filters with the programmable pipeline. It sounds like the issue is not the min/mag filter, but with how you're handling texture clamping/wrapping. Either that or your textures have gray in them, which you probably don't want.
To set up texture clamping, you can do the following:
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
This will cause any pixels sampled from outside the texture to return the same color as the nearest pixel within the texture to that sample location.
As the other answers and comments alread pointed out, the texture sampling states will effect both the fixed function pipeline and the programmable pipeline in the same ways. I'd just like to add that in shaders, you can also completely bypass the sampling and use the GLSL texelFetch() functions where you can directly access the unfiltered texels - which will basically look like GL_NEAREST filtering. You will also lose the wrapping functionality and hve to use unnormalized integer texture coords, so this is probably not what you want in that scenario, though.

Mipmaps and Nearest filtering result in darker image

I am loading images into OpenGL app.Usually I am using Linear filtering but now testing nearest I found the resulting image is significantly darker than the original one.Btw,it also seems to me that the linear filtering causes some brightness loose too.Here are examples:
Linear filtering :
Nearest filtering :
Original image:
Now, I am setting mipmaps levels (to 4 ).I found that when not using mipmaps the original brightness is intact.What can be the problem?Is it related to gamma correction?
Here is the code for image load and mipmap generation:
ILinfo imageInfo;
iluGetImageInfo(&imageInfo);
iluFlipImage();
if (imageInfo.Format == IL_RGB)
{
ilConvertImage(IL_BGRA, IL_UNSIGNED_BYTE);
}
else if (imageInfo.Format == IL_RGBA)
{
ilConvertImage(IL_BGRA, IL_UNSIGNED_BYTE);
}
iluGetImageInfo(&imageInfo);
glPixelStorei(GL_PACK_ALIGNMENT, 1);
glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
glGenTextures(1, &textureName);
glBindTexture(GL_TEXTURE_2D, textureName);
glTexStorage2D(GL_TEXTURE_2D,numMipMapLevels,GL_RGBA8,imageInfo.Width,imageInfo.Height);
glTexSubImage2D(GL_TEXTURE_2D,0,0,0,imageInfo.Width,imageInfo.Height,GL_BGRA,GL_UNSIGNED_BYTE,imageInfo.Data);
/* ==================================== */
// Trilinear filtering by default
if(smooth){
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glGenerateMipmap(GL_TEXTURE_2D);
}else{
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST_MIPMAP_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
}
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
I am also running MSAA pass in a custom FBO but it looks to be irrelevant to the issue as I tested also with MSAA turned off and the same problem persists.
From your code it looks like you create a mipmapped texture (with 4 mipmap levels as you say) but then only set the image for the first level. This means all the other level's images are undefined. When then using GL_NEAREST_MIPMAP_LINEAR, it will access the two mipmap levels that best fit the pixel-texel-ratio (the MIPMAP_LINEAR-part) and then pick a single nearest texel from each level (the NEAREST-part) and interpolate those.
From your image it looks like the unspecified mipmap levels are just black, so you get an interpolation between the texture color and black, thus a darkened texture (well, they could actually contain anything and the texturing shouldn't even work since the texture is incomplete, but maybe immutable storage behaves different in this regard). When not using mipmaps (thus only creating a single level with glTexStorage), there will only be a single level used in the filtering (even if using a mipmapped filter), which of course has a valid image.
If you intend to use some kind of mipmapping, then you should actually set the texture image for each and every mipmap level (or set the top-level image and do a glGenerateMipmap call afterwards). If you just wanted to use real nearest neighbour filtering, then just use GL_NEAREST (I've never actually seen much practical use for all the other mipmap filters except for the real trilinear filter GL_LINEAR_MIPMAP_LINEAR).

How to check for mip-map availability in OpenGL?

Recently I bumped into a problem where my OpenGL program would not render textures correctly on a 2-year-old Lenovo laptop with an nVidia Quadro 140 card. It runs OpenGL 2.1.2, and GLSL 1.20, but when I turned on mip-mapping, the whole screen is black, with no warnings or errors.
This is my texture filter code:
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_GENERATE_MIPMAP, GL_TRUE);
After 40 minutes of fiddling around, I found out mip-mapping was the problem. Turning it off fixed it:
// glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
// glTexParameteri(GL_TEXTURE_2D, GL_GENERATE_MIPMAP, GL_TRUE);
I get a lot of aliasing, but at least the program is visible and runs fine.
Finally, two questions:
What's the best or standard way to check if mip-mapping is available on a machine, aside from checking OpenGL versions?
If mip-mapping is not available, what's the best work-around to avoid aliasing?
Edit:
Here's the question I'm really after:
How do you programmactically check if mip-mapping is available on a platform?
OpenGL 1.4 is required for support for automatic mipmap generation (using GL_GENERATE_MIPMAP), texturing using mipmaps should be available on virtually every graphic card out there.
I guess the problem might not be
missing mipmapping functionality
but maybe the generation of the
mipmaps. Just for debugging
purposes, can you try generating the
mipmaps using gluBuild2DMipmaps?
If this problem is specific to this graphic card/configuration, checking for updated drivers might be a good idea, too.
A good method to get rid of aliasing is multisampling.
If no mip-mapping is available, your probably don't want to avoid the aliasing. A platform without mip-mapping is usually not worth the trouble.
However, if you really really want to avoid aliasing, you could approximate per-fragment mip-mapping by per-primitive mip-mapping. First you have to generate and upload all mip-map-levels by hand into separate textures. Before rendering a triangle, compute the appropriate mip-map level on the cpu and use the texture corresponding to this mip-map level. To avoid artifacts on large primitives, subdivide them.
OpenGL requires that all texture formats that are part of the Core support mip-mapping. As such there is no specific way to ask whether a texture format supports mip-mapping.
Some extensions, when first introduced, were relaxing this constraint for new formats (most notably, non-power of two textures, and float textures). The extension itself made it clear that the mip-mapping was not supported at all. But before those got integrated into the core, the requirement for mip-mapping support was reinstated.
I would be very surprised if your implementation really didn't support mipmaps. More likely you are missing one or more levels, this is something that can easily happen by accident.
If you only miss as much as a single mip level and enable mipmapping, you get just what you have, black. Be sure that either all levels down to 1x1 are defined, or set the GL_TEXTURE_MAX_LEVEL TexParameter accordingly.