Previous | Next --- Slide 49 of 64
Back to Lecture Thumbnails

How do you determine which method of rasterizing a line is better? I personally think they all look around the same?


Is the diamond rule computationally efficient? It seems like it would require a lot of processing power for a very simple problem.


How are the GPUs efficient for computer graphics?


I've heard a technique called anti-aliasing. Does this technique use an algorithm similar to the diamond rule?


Is this only used for lines on modern GPU's? And what's the advantage (performance or appearance-wise) for doing so?


Is there a "circle rule" where they used circles instead of diamonds? What are the advantages/disadvantages of using diamonds instead?


Is there a specific reason of choosing "diamond" as the criteria here?


How would you be able to tell if the line passes through the diamonds? Would you have to encode coordinates for each diamond on each pixel, keeping estimated "line thickness" in mind, or is there a different approach?


How exactly do modern GPU's compute/store this "diamond" information about the pixels?


Are there some ways to quantify how much better one rasterization technique is compared to another kind?


How is line rendering using the diamond rule parallelized on GPUs? Is the line split into sections drawn in parallel or is parallelism not that important for an individual line?


Is the diamond method only used for 0-thickness lines?


Have there been attempts at improving this "diamond" technique or is it the most efficient we have so far?


Since the line divides every pixel that it touches into two parts, can we calculate the ratio of the area of the two parts and use that as a criteria to decide whether to light up a pixel?


Does everything use the diamond rule or are there other rules that are used?


Are there other rules (i.e circles instead of diamonds) that work better for different orientations / styles of lines and pictures, and are there easy ways to discern as to which rule should be used for a specific image?


What are some examples of other rules that work better for different criteria? Would it be possible to look at some analysis of that? Also, it seems to me that the diamond rule is chosen partly for reasons of computational efficiency on a GPU (just 4 sides of the diamond to intersect against) rather than another shape which would entail more complex operations.


How exactly do we divide up the pixel in half or determine the diamond shape for our computations? And how do we create this non-zero thickness line to actually do these computations?


I am also interested in, is there a specific reason that this diamond rule can help GPU acceleration, or other rules, say circle or hexagon, is equivalent in speed? Then what's the reason for using these diamonds?


How different can a picture look based on the rasterization technique used to light up and color the objects in the picture? Are these differences noticeable to the naked eye?


Are there any other shapes tried to get a smoother line?


How we implement the diamond rule?


Is there any efficient method to store all those information?


This looks like it would be vulnerable to aliasing on horizontal lines, which are fairly common, since they could be thin but light up two whole rows. Is there a way around that?