Can depth pixel numbers be extracted from THREE.WebGLRenderer, similar to the .readRenderTargetPixels functionality? Basically, is there an update to this question. My starting point is Three.js r80. Normalized values are fine if I can also convert to distances.
Related methods:
I see that WebGL's gl.readPixels does not support gl.DEPTH_COMPONENT like OpenGL's .glReadPixels does.
THREE.WebGLRenderTarget does support a .depthTexture via THREE.WebGLRenderer's WEBGL_depth_texture extension. Although THREE.DepthTexture does not contain .image.data like THREE.DataTexture does.
I also see that THREE.WebGLShadowMap uses .renderBufferDirect with a THREE.MeshDepthMaterial.
Data types:
A non-rendered canvas, can use .getContext('2d') with .getImageData(x,y,w,h).data for the topToBottom pixels as a Uint8ClampedArray.
For a rendered canvas, render() uses getContext('webgl') and contexts may only be queried once, so getImageData cannot be used.
Instead render to a target and use .readRenderTargetPixels(...myArrToCopyInto...) to access (copy out) the bottomToTop pixels in your Uint8Array.
Any canvas can use .toDataURL("image/png") to return a String in the pattern "data:image/png;base64,theBase64PixelData".
You can't directly get the content of the FrameBuffer's depth attachment using readPixels. Whether it's a RenderBuffer or a (Depth) Texture.
You have to write depth data in the color attachment.
You can render your scene using MeshDepthMaterial, like shadow mapping technic. You ends up with the depth RGBA encoded in the color attachment. You can get it using readPixels (still RGBA encoded). It mean you have to render your scene twice, one for the depth and one to display the scene on screen.
If the depth you want match what you show on screen (same camera/point of view) you can use WEBGL_depth_texture to render depth and display in one single render loop. It can be faster if your scene contains lots of objects/materials.
Finally, if your hardware support OES_texture_float, you should be able to draw depth data to a LUMINANCE/FLOAT texture instead of RGBA. This way you can directly get floating point depth data and skip a costly decoding process in js.
Related
I am attempting to create a view for Android whose background appears to blur the view's content for which it is on top. This is nothing new and has been done before. I based my implementation on what was done here: Dimezis/BlurView.
The approach uses the pre-draw event from the view tree observed to draw a view to an internal canvas. The canvas is backed by a bitmap. A blur is applied to the bitmap before it is drawn to the canvas passed to BlurView's draw method. This approach works well for all standard views/controls and is a common method used to achieve the blur view effect.
However, it does not handle Skia-based drawings that may be on the view that is blurred.
Skia, via SKCanvasView, is used heavily for controls within the App so this is kind of a deal breaker if I cannot find a solution. The issue is very odd. Anything that is drawn on the canvas view appears to be scaled and translated when drawn to the internal canvas.
Screen-Shot - blur label v. blur SKCanvasView
The screen-shot show the difference in results from blurring a label with text v. blurring a red circle drawn on an SKCanvasView.
For reference, I've posted a sample project on GitHub. It can be found here: jaredballen/BlurView
I'd really appreciate any input that can be shared.
When you invoke _rootView?.Draw(_internalCanvas); the circle SKCanvasView is internally rendering itsself using _internalCanvas size (same as _blurView size) resulting in demonstrated behavour. My guess this could be solved by making _internalCanvas of the same size as the _rootView, rendering root view as blurred, then clipping and translating the result inside the smaller _blurView.
I want to render some graphics, texts and shapes to a bitmap using a canvas-like API like nanovega in D on my server.
I know how to create an arsd window with OpenGL context and render to it, (as per documentation) but is it also possible rendering to a headless context or even directly to draw to some memory buffer? (Because I don't think my server has OpenGL available and would probably need to use some software renderer)
Mainly I wanted to use gradients, texts, shapes, rounded corners, images and masks and render all of that to an image. I know that nanovega implements all of the rendering parts of that, so I would like to keep using it.
Recently I am writting game editor in my project. I want to implement a editor which has four viewport like 3ds max or other 3D software.
So, how to use opengles2 to render context on multi-window?
You can usually have multiple views with each having its own frame buffer. In this case all you need to do is bind the correct frame buffer before drawing to each of the views. You might also need to have different contexts for each view and setting them as current before drawing (also before binding the frame buffer). If you need multiple contexts you will need to find a way to share resources between them though.
Another approach is having a single view and simply using glViewport to draw to different parts. In this case you need to set glViewport for a specific part, setting ortho or frustum (if view segments are of different size) and that is it. For instance if you split the view with buffer having dimensions bWidth and bHeight into 4 equal rectangles and you want to refresh the top right:
glViewport(bWidth*.5f, .0f, bWidth*.5f, bWidth*.5f);
glOrthof(.0f, bWidth*.5f, bHeight*.5f, .0f, .1, 1.0); //same for each in this case
//do all the drawing
and when you are finished with all you want to update just present the frame buffer.
I am trying to create a cube using three.js for a project. I need to add text to vertices and at different points inside the cube. Any idea how this can be done?
For some basic code examples of using Sprite objects in Three.js, check out:
http://stemkoski.github.com/Three.js/Sprites.html
And for an easy way to create images that contain text to use as your sprite textures, check out the sample code at:
http://stemkoski.github.com/Three.js/Texture-From-Canvas.html
I think a combination of these two ideas will achieve what you are looking to do.
If you want to have label-style texts, so that the text begins at a specific point, but is always oriented with the camera and easily readable no matter the camera position, you can use sprites. (example of canvas-created text label sprites: http://i.imgur.com/e9I68xD.jpg - here they are rendered on a separate pass to that they are never obscured by the scene but you can do it on the same pass)
If that's what you are looking for, I'd suggest first checking the Sprites examples, and learn to attach some static image as a sprite to correct position in the scene. After you get that working, you modify the code so that you generate the text to an image canvas using standard Javascript Canvas functions, and using that image as the sprite.
Within my current MATLAB GUI project, i have two axes objects. The first is used by the workaround "uibutton" (I don't use GUIDE) in order to display a LaTeX-formula (as far as I know, only axes labels are capable to use LaTeX whereas normal static text fields aren't...). The other axes object is used to actually plot a 3D-function.
The program has the following steps:
the first axes creats the LATEX-formula (e.g. f(x)=).
User enters a function in the edit field after the LaTeX-formula (e.g. f(x)=a+b).
User presses a "plot"-button.
3D-function is plotted in the second axes object.
Problem:
As soon as the 3D-function is plotted, the nicely rendered LaTeX-formular becomes crisp. Is there any way to prevent this from happening?
http://i42.tinypic.com/348pq2u.png (See picture for problem demonstration)
Check your figure properties before and after you draw the 3D plot
get(gcf, 'renderer')
My guess is that plotting the 3D function changes the renderer from the default ("painters") to another (likely OpenGL). Matlab's Latex rendering does not seem to play well with zbuffer or OpenGL (these produce bitmaps rather than line art).
You may be stuck if painters can't render your 3D graphics properly, but you can try to force it by setting the renderer manually back to painters
set(gcf, 'renderer', 'painters')