https://www.youtube.com/watch?v=gCWhWBtu0LA
The NES is connected to the graphics memory on the cartridge via groups of signals called buses. There's the Address Bus and the Data Bus. When the NES wants to access graphics memory on the cartridge, it puts the memory address it wants in the Address Bus. The memory chip then grabs the data at this address and transfers it to the NES via the Data Bus.
Now, you'd think in order to emulate a memory chip, you'd have to watch the Address Bus to know what data to send back, which would be quite complicated to do. And that's true in general. However, the NES accesses memory in the exact same order each time for each frame of video. As we can predict exactly what data it will want at any given time, we can just feed it data in that exact order and ignore the Address Bus entirely. This means that our logic circuitry and software can be much simpler.
There's another benefit to ignoring the Address Bus, and it has to do with the way NES graphics work. The NES screen, which is 256×240 pixels, is divided into 16×16 pixel sections called attribute areas.
Each attribute area can only have four different colors displaying at one time, which makes for a pretty bad color resolution. The colors that each attribute can display are stored in an area of memory called the Attribute Table.
When the NES wants to check which colors a particular attribute can display, it reads that part of memory. But the NES actually does this 32 times per attribute area, once for each 8×1 pixel section of the attribute.
The NES looks at the same memory bus address for all of these 8×1 pixel sections, and it's expecting to get the same value back each time on the Data Bus.
But since we're ignoring the Address Bus, we can feed it a different set of colors for each section, thus fooling the NES into giving us 8×1 pixel attribute areas instead of the normal 16×16 pixel.
And this results in a much better color resolution than NES games can normally achieve.
However, even with these techniques, we're limited to only 13 different colors on screen at once.
We can increase the perceived number of colors with a process called dithering, which is where you deliberately apply noise to the picture. This stops the banding effect you get on color gradients but gives the picture a kind of speckly appearance.
There are many things that could be done to increase the video quality, such as dynamic palette generation or using sprites to add additional colors to certain parts of the screen. Or even just using a better dithering algorithm, such as one of the ones invented by Joel Yliluoma.
