If your image is RGB24 then VideoReader always reads and outputs the entire data. If your image was indexed then you could specify 'native' to get back a struct that had the indexing information and the colormap and so would be slightly faster.
VideoReader always returns the entire frame; there is no crop option (at least up to R2015b).
The methods of setting cropping and format to return data only work for live video, not reading from a file.
Best Answer