If the matrix had indeed rendered a complete image of what the virtual world looked like, then that image would then be preprocessed by the part of the brain responsible for visual inputs.
The amounts of data passed on to the conscious part of the brain is much smaller. But that doesn't mean the matrix could have cut any corners in the rendering process. If the rendering didn't produce a perfect image, then the preprocessing done by the visual part of the brain would pick up the rendering artifacts and alert the conscious part of the brain to the rendering artifacts and not have it pay as much attention to everything else.
In the movie there is evidence that this was solved by completely bypassing the part of the brain performing this preprocessing of visual inputs. It is more likely that instead of going from simulated model to images to processed images, the matrix went directly from simulated model to processed images. The processed images of course have to be injected at a different place in the brain than the preprocessed images would have been.
In the movie it is explained that turning the model data into images is computationally expensive, and hence is rarely done. Instead when inspecting the model data on a computer screen it is just shown as a series of symbols closely matching the model data.
The computational cost was mentioned in the context of what the computers on the ship could do. But it seems quite reasonable that the matrix wouldn't waste computational resources on this unneeded computation. Also if the matrix had actually done the computation, we would have expected that on the ship they could simply have switched their computer monitors to display these rendered images instead of the model data.
This means that in the movie, whenever you are looking at images from within the matrix, you are looking at images that do not exist in-world. They represent the step that was actually left out from the computation.
Obviously the images had to be rendered for the movie, because otherwise watching the movie would have required years of training in decoding the symbols on screen or a plug directly into your brain.
Before the question then comes up about why it was even possible to render the movie, if the rendering was so expensive, it should be pointed out that to produce the movie only rendering from a single point of view was needed and it could be done offline. Should the same have been done inside the matrix it would have had to be real-time with a different point of view rendered for every human.
So if Neo were to look into a mirror while inside the matrix. He doesn't see an image of what he looks like. He simply see a tag saying that is me plus possibly tags indicating any parts of his body looking in an unexpected way, for example if a bruise had suddenly turned up in his face that would be tagged separately.
It is only once he leaves the matrix that actual images starts getting attached to the tag. If Neo were to look at his own hand in the real world and turn it around in front of his eyes, his brain would realize that since this hand was turning as he was turning his hand, then that was his own hand, and the image would be remembered for future comparisons.