PyOpenGL Texture_3D and Numpy not rendering

Question

I'm trying to do a simple rendering loading data from TIFF images into numpy matrices and making a 3D texture. Tiff images are intensity images, I was hoping to remap this intensity image as a luminance or alphacolor image.

While I think I understand the process of GL_Texture_3D, with no errors running the code I still obtain no visual result. I certainly am doing something simple very wrong. Could anyone help me understand my errors?

This is a short version of the code [corrected]:

Edit 2: trying to cast types correctly: as suggested by Reto Koradi I might be facing a casting issue. So I convert numpy array to uint8 making sure values are different than all zeroes and pass it to the channel_RGBA_buffer uint8 arrray. Still produces a black window.

Edit 3: using the Glumpy approach: reading the function "textures" of Glumpy ( https://code.google.com/p/glumpy/source/browse/glumpy/image/texture.py?r=8cbc2aa2b277c07b59fba964edd327370a8e8091 ) I changed the target GL format GL_RGBA to GL_RGBA16 and added glBindTexture( GL_TEXTURE_3D, 0 ) after calling glTexImage3D. Still unsure as to why 0 and not the texture.

Edit 4: fixed indentation of 2 lines to make example executable

Result is still a black window but when closing this window, for an instant, I see the data appearing. Seems like I'm putting an extra layer.

import sys

import numpy
from OpenGL.GL import *
from OpenGL.GL.shaders import *
from OpenGL.arrays import vbo
from OpenGL.GLU import *
from OpenGL.GLUT import *
from OpenGL.GLUT.freeglut import *




class test_TIFF_render:

    def __init__(self):
        self.something = True

    def load_texture(self):

        global texture


        #read file as array, for simplicity I will just make a random numpy 3D matrix 
        tiff_mean = numpy.random.rand(3,300,300)
        tiff_mean = tiff_mean*1000
        tiff_mean = tiff_mean.astype('uint8')

        shape_array = tiff_mean.shape
        shape_array = tiff_mean.shape
        #we have the data now, let's set the texture
        tiff_mean = tiff_mean.reshape(tiff_mean.shape[1]*tiff_mean.shape[2]*tiff_mean.shape[0])
        texture = glGenTextures(1)
        glBindTexture(GL_TEXTURE_3D,texture)
        glTexEnvi(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE)
        glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S,GL_CLAMP_TO_BORDER)
        glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T,GL_CLAMP_TO_BORDER)
        glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R,GL_CLAMP_TO_BORDER)
        glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR)
        glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_LINEAR)

        #now we need to convert the intensity data organized in 3D matrix into an RGBA buffer array (pixel1R, pixel1G, pixel1B, pixel1A, pixel2R....)
        channel_RGBA_buffer = numpy.zeros(tiff_mean.shape[0]*4, numpy.uint8)

        for i in range(0,tiff_mean.shape[0]):
            channel_RGBA_buffer[i*4] = tiff_mean[i]     #R
            channel_RGBA_buffer[i*4+1] = tiff_mean[i]   #G
            channel_RGBA_buffer[i*4+2] = tiff_mean[i]   #B
            channel_RGBA_buffer[i*4+3] = tiff_mean[i]   #A
            if numpy.mod(i,100000)==0:
                print('count %d',i)

        
        glTexImage3D(GL_TEXTURE_3D, 0, GL_RGBA16, shape_array[1], shape_array[2], shape_array[0], 0, GL_RGBA, GL_UNSIGNED_BYTE, channel_RGBA_buffer)
        glBindTexture( GL_TEXTURE_3D, 0 )

    def display(self):

        glClear( GL_COLOR_BUFFER_BIT  | GL_DEPTH_BUFFER_BIT )

        glEnable( GL_ALPHA_TEST )
        glAlphaFunc( GL_GREATER, 0.03)

        glEnable(GL_BLEND)
        glBlendFunc( GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA )

        glMatrixMode( GL_TEXTURE )
        glLoadIdentity()

        

        glEnable(GL_TEXTURE_3D);
        glBindTexture( GL_TEXTURE_3D, texture);
        dOrthoSize = 1
        for Indx in self.my_range(-1,1,0.01):
            TexIndex = round(Indx,2)
            glBegin(GL_QUADS)
            glTexCoord3f(0.0, 0.0, (TexIndex+1.0)/2.0)
            glVertex3f(-dOrthoSize,-dOrthoSize,TexIndex)
            glTexCoord3f(1.0, 0.0, (TexIndex+1.0)/2.0)
            glVertex3f(dOrthoSize,-dOrthoSize,TexIndex)
            glTexCoord3f(1.0, 1.0, (TexIndex+1.0)/2.0)
            glVertex3f(dOrthoSize,dOrthoSize,TexIndex)
            glTexCoord3f(0.0, 1.0, (TexIndex+1.0)/2.0)
            glVertex3f(-dOrthoSize,dOrthoSize,TexIndex);
            glEnd()

        glBindTexture( GL_TEXTURE_3D, texture )

    def my_range(self,start, end, step):
        while start <= end:
            yield start
            start += step

glutInit(sys.argv)
glutInitDisplayMode(GLUT_RGBA)

glutInitWindowSize(300,300)
window = glutCreateWindow(b"I really hope this works")


test111 = test_TIFF_render()
test111.load_texture()

glutDisplayFunc(test111.display)
glutMainLoop()

I am running Win7 64bit with Python3.3 .

Any reason in particular you are using GL_CLAMP_TO_BORDER yet have not defined a border color anywhere? Did you mean to use GL_CLAMP_TO_EDGE by chance? — Andon M. Coleman
– Andon M. Coleman, Commented Jun 10, 2014 at 2:00
I was naively following/porting this example to python ( codeproject.com/Articles/352270/… ) — HeoN
– HeoN, Commented Jun 10, 2014 at 18:04
Ah, okay. That is actually perfectly valid then and you can ignore my comment :) The default border color is black, so this will produce black texels for any coordinates sampled outside the range [0,1]. Assuming empty space is supposed to be black, that behavior is acceptable. GL_CLAMP_TO_EDGE would do something entirely different, it would sort of take the edge color and stretch it out to infinity. Ther is a nice visual summary of the wrap modes here (border color=red in this case). — Andon M. Coleman
– Andon M. Coleman, Commented Jun 10, 2014 at 18:11

Reto Koradi · Accepted Answer · 2014-06-11 05:43:12Z

1

I haven't used numpy, but the way I read the documentation (http://docs.scipy.org/doc/numpy/reference/generated/numpy.zeros.html), numpy.zeros() produces an array of float64 values by default. You then pass GL_INT as the type argument for glTexImage3D(). It looks like we have a type mismatch here.

The typical data type to use for texture data is unsigned bytes. So I think it should look something like this:

channel_RGBA_buffer = numpy.zeros(tiff_mean.shape[0]*4, numpy.uint8)
...
glTexImage3D(GL_TEXTURE_3D, 0, GL_RGBA, shape_array[1], shape_array[2], shape_array[0], 0,
             GL_RGBA, GL_UNSIGNED_BYTE, channel_RGBA_buffer)

Edit, updating for new version of code: It looks like you know generate random float values between 0.0 and 1.0, cast them to byte (uint8) values, and then use that for the texture data. I believe this will result in all your byte values being 0, or maybe partly 1 depending on how the rounding works. But the value of the byte range is 0-255. You probably want to multiply by 255.0 when converting your random floats to the texture data:

channel_RGBA_buffer[i*4] = 255.0 * tiff_mean[i]     #R
channel_RGBA_buffer[i*4+1] = 255.0 * tiff_mean[i]   #G
channel_RGBA_buffer[i*4+2] = 255.0 * tiff_mean[i]   #B
channel_RGBA_buffer[i*4+3] = 255.0 * tiff_mean[i]   #A

edited Jun 11, 2014 at 5:43

answered Jun 10, 2014 at 3:23

Reto Koradi

54.7k8 gold badges100 silver badges144 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

HeoN Over a year ago

while I agree that is a major flaw in my code, there must be something else because this does not solve the problem. Does it have to be uint8? I thought I could use int16. I corrected the code in the main post but still no result.

Reto Koradi Over a year ago

You can have textures with 16-bit components. But you have to be careful that the format of the data you pass to glTexImage3D() matches the format parameter.

Reto Koradi Over a year ago

I updated the answer based on your latest code. Frankly, I'm only a very casual Python user, so I'm not always sure how it handles data types and conversions.

Collectives™ on Stack Overflow

PyOpenGL Texture_3D and Numpy not rendering

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related