We are using simple differential centrifugation for bioenergetics studies (see: http://link.springer.com/article/10.1007/s00227-013-2273-x). Its' easy and fast. However, in case you need highly purified mitochondria you may want to perform a Percoll gradient centrifugation step. The protocols developed for rat liver mitochondria should work well on fishy too (see: http://www.biochemj.org/bj/204/0731/2040731.pdf).

As Michael suggested we always performed differential centrifugation for liver mitochonria in past. If you wants to get purified mitochondria and have not perform percoll gradient before...you can also use ficoll gradient (7.5% & 10%). I prefer ficoll method because you get purified mitochondria as pellet versus at inter-phase in Percoll method.

Keep in mind that isolated mitochondria will also have 'fragments' of other organelles closely bound to them even when purified. Therefore, you have to allow for this when performing proteomic studies. It is worth checking your preparations for marker enzymes of other cellular membrane systems routinely, to assess the purity of each batch.