I recently read a great article on copyright law and translation by Sarah Kolb for the Writers 101 site.
The article discusses the idea of translating a book protected by a copyright. The translation of a book is a huge undertaking. Literary translation is a very specialized form of translation that is typically done by linguists who could be considered authors in their own right. It would follow that an author would be pleased to have their work introduced in new markets. But the article points out how that would potentially limit the author in benefiting from sales in secondary markets and the decision to release the book in other markets should rest completely with the holder of the copyright.
The author goes on to explain that even if the translation is distributed for free the ability to sell in that market would certainly be impaired by a free translation being circulated. It seems that the principal question is one of distribution. I would imagine that if a book is translated for own personal use and never shared it would not violate copyright law. Now, what if a machine translation tool in the public domain is used to create that translation? I wrote about Google translate as a form of plagiarism in a blog post in 2011. Seth Godin recently posted on the idea of fair use which is a central principle in determining violation of copyright laws.
As the quality of machine translation improves it will be interesting to see how seriously MT output will be taken in the marketplace. As of now literary MT is really not an option. The quality simply is not acceptable yet.