Antialiasing

Aliasing: Unwanted Image artifacts
An interesting aspect of image rendering lies in the elimination or, if not possible, of the attenuation of image artifacts; i.e. visible defects in the image. These defects can range from unrealistic lighting effects to awkward looking reflections. But in a digital pixel-based image, none is more obvious than what is referred to as aliasing. Aliasing, refers to the problem that a digital image being built out of discrete elements, here pixels, will have certain of those elements which should have been distinct, copy each other. For the pixels in an image, this aliasing artifact is referred to as "jaggies". The jaggies are mostly found in areas where there is a sudden change in light intensity or colour,  creating instead of a smooth line, a "jagged" edge.


Antialiasing
Antialiasing, as its name suggest, is the set of methods used to remove or reduce pixel related artifacts. There are many algorithms that can be used to alleviate this problem: Supersampling and Stochastic (or random) ray sampling are described in Glassner's[1].


Supersampling
The method we have chosen to experiment with will be supersampling. It is the simplest, and the most adaptable to our present implementation. The concept is very simple: Instead of having one ray per pixel which could create a hard border between the edge of a sphere and its background, we fire multiple rays within the viewing window's pixel area. We then average out the intensities for the particular pixel and send that average intensity to the image buffer. The result is a much smoother image, 


The ViewPoint class possesses a member method antialiasing_vectors that adds into the pixel area of the viewing window 4 evenly distributed coordinates where a ray will be sent to. This is easily done by changing the area's offsets.

Supersampling vs single ray sampling

Adding more rays will obviously augment rendering time; But since performance is not the issue, the results are worth the wait. We can easily compare an antialiased image to its non-antialiased version, by inserting the 'a' command line argument when launching the ray tracer application. Here are comparison results:

Rendered image with antialiasing at OFF

Same image with antialiasing

The antialiasing is best seen here with close-ups of the previous images:

Antialising is OFF

Antialiasing is ON

There are more efficient ways of reducing the aliasing effects than just by augmenting the number of rays per pixel area. Stochastic antialiasing applies a random spreading of the rays inside the pixel area, which will have similar effects as supersampling, but, depending on the number of extra rays involved in the supersampling process, it could potentially be less taxing, while achieving similar results.