Week 11 - Motion Blur

On Monday a lot of the studios got to show off their games to the rest of the class. Everybody's games are starting to come together and look really great, a huge step up from first semester.

Above is a screenshot of the new lighting system in our game. By next week i'll have screenshots of our entire first level instead of just one model. 

Motion Blur:

Motion blur is an effect that is caused when the image being recorded or rendered changes during the recording of a single frame. This can be caused either by rapid movement or long exposure. 

There are many ways of implementing motion blur. There is an easy way that involves the use of the accumulation buffer. The downside to this method is that it requires rendering each frame multiple times, if you want a lot of blur or have a lot of geometry that can cause slow downs.

- draw the frame

- load the accumulation buffer with a portion of the current frame

- loop and draw the last n frames and accumulate.

- display the final scene

The more modern way or effective way of doing motion blur involves the use of motion vectors. To do motion blur with motion vectors you need to calculate each pixel's screen space velocity then that velocity is used to do the blur. The calculation of this vector is done in fragment shader on a per-pixel basis. 

Week 10 - Depth of Field

This week we learned about depth of field and had an in-class competition on friday.

Depth of field is an effect that causes objects that are out of focus to appear blurry.
Computer graphics uses the pinhole camera model which results in perfectly sharp images. The pinhole camera model only lets a single ray through. Real cameras use lenses with finite dimensions which is what causes depth of field.

Depth of Field Implementation:
- Use destination alpha channel to store per-pixel depth and blurriness information.
- Use fragment shader for post-processing
- Downsample and pre-blur the image
- Use variable size filter kernel to approximate circle of confusion
- Blend between original and pre-blurred image for better image quality
- Take measures to prevent "leaking" sharp foreground into blurry background
- We pass the camera distance of three planes to scene shaders
     - Focal plane: points on this plane are in focus
     - Near plane: Everything closer than this is fully blurred
     - Far plane: Everything beyond the far plane is fully blurred

Week 9 - Lighting and Deferred lighting

This week we watched a conference video about God of War's lighting system and we also looked at deferred lighting.

 During the presentation they explained the process behind their shadows:

ZPrePass -> Cascade 2 -> WB Shadow Map -> Cascade 1 -> WB Shadow Map -> Cascade 0 *

* -> WB Shadow Map -> Opaque -> Transparent+Effects+UI+Flip

I'm really interested in learning more about their use of the white buffer and just the uses of the white buffer in general.

Our second lecture this week was on deferred lighting.

Deferred lighting is a screen-space lighting technique where the lighting is postponed or deferred until the second pass hence the name deferred lighting or shading.

One of the advantages of deferred lighting is lighting is now based on number of lights instead of the actual geometry allowing you to have a lot more lights in your scene. Some downsides to deferred lighting is that its difficult to do antialiasing, transparent objects still need to be done separately, hard to use multiple materials, memory bandwidth, and shadows are still seperate.

Week 8 - Midterm and Tidbits

On Monday we had our midterm for intermediate computer graphics. I think it was a pretty good midterm, it wasn't too difficult but also it wasn't too easy. 

During our lecture this week Dr. Hogue talked about little tidbits to add on to what we've already learned. First we talked about colours and the different ways of representing colour.  

We also talked about thresholding. Thresholding is a image processing effect that basically turns an image into a binary image (black and white). Thresholding is useful in technologies like QR code cameras

We also looked at several different convolution kernels like image sharpening, blurring and motion blur.

Week 6 - Brutal Legend and Midterm review

We didn't learn anything new this week but we did watch a video about the making of Brutal Legend to show how what we learned is applied in popular games. Our second lecture this week was just a review for our midterm after reading week. 

Overall the video was really interesting to watch and the developers made some really smart choices when developing Brutal Legend. One thing that really surprised me was how they chose to render the sky. They chose to make their sky one giant particle instead of using a traditional sky-box. They also talked a lot about how they did particle rendering and lighting. 

Week 5 - Lighting and Shadow Maps

This week we talked about global illumination and shadow mapping. 

     Global illumination is a general name for a group of algorithms used to give more realistic lighting to a 3D scene. These algorithm's allow objects to be lit not only from light rays directly from the light source but also by rays that have bounced off of other surfaces. As seen in the image above the white surfaces are tinted with colour cause by the light rays that have bounced off of the coloured green and red walls. 

     Ray tracing is a technique that can be used to achieve more realistic lighting in a scene. Ray tracing is basically trying to simulate rays of light. Ray tracing is done by tracing rays of light from a light source and checking to see how it interacts with the objects in the scene and render accordingly.

 

Shadows:

     Shadows are caused by absence of light in an area. In a 3D scene this is usually because there is an object in-between the light source and the object creating an obstruction for the light.

Shadow Maps:

The algorithm for shadow mapping that we went over in class required 2 rendering passes. During the first pass you would render the scene from the light's viewpoint, draw blockers and store nearest z in the z-buffer and the resulting depth buffer is the shadow map. The second pass is done from the observer's viewpoint and you compare the depth of each pixel against the shadow map to determine which pixels are in shadows.

Week 4 - Fullscreen Effects/ PostProcessing

This week we learned about full screen effects and post processing effects like blur and bloom. 

Blur: 

Blurring is done by applying a filter to the texture of your frame where each frame is assigned a weighting. 

In the image above each pixel is equally weighted.

Gaussian blurring is done in a similar way except each source pixel is not equally weighted, pixels are weighted higher in the center. As a result, bigger the window the stronger the blurring.

HDR/Bloom:

HDR and Bloom are post processing effects that are actually quite simple to do, they are done by doing multiple pass-throughs before displaying the final image.

     1. Render 3D scene to offscreen framebuffer
     2. Highlight bright areas(tone-mapping)
     3. Apply a Gaussian blur to highlighted areas
     4. The final image is equal to the blurred frame + the initial 3D scene

Above is an example of HDR and bloom in a video game, in this case it is The Legend of Zelda : Wind Waker HD remake. 

GDW:

We are planning on adding some postprocessing effects into our game and currently we are working on our framework to make these effects easier to implement.