python multiprocessing performance in opencv image processing

edit

I am writing a program with Opencv in python which takes a video, does some filters and then goes into a detectmultiscale. I am searching a way to increase fps processed and I split in 2 processes this 2 operations (filtering and detecting) using queue to put and get the frame.

Now the filtering goes at double speed against the detection but the overall fps remains the same. I am asking why and how can I increase performance.