I have some footage from my video camera taken in 1080i-AVCHD (.MTS file extension). Loading the clips in the VSE goes fine. When I set the framerate to 50 before loading the clips the sound is as long as the video. If I set the frame rate to 25 the video clip has double length of the sound. After loading a number of clips and adding effects I change the frame rate from 50 to 25 and enable de de-interlace filter on all clips before rendering the whole with the h264 codec presets. The result is OK.
My question : is this a normal workflow for this application or do I miss something that could have a better quality of the rendered video ? (although the quality I have is already good)
It looks strange to me that there is no separate setting for the fps of the input footage and the fps of the rendering.
Sounds like you’re done fine and you’re happy with the results so it’s a perfectly good workflow. 
Only comment really on your process is that it’s not compulsory to deinterlace, it will very possibly reduce the quality in itself. Blender’s deinterlace method is ok, code ported from VirtualDub I think.
With regard to quality, I think you can make improvements over what you have, probably but it’s whether you really want to go the extra.
If you’re interested there are some notes on the blog in my sig.
Thanks for your reply yellow. I have read about Avisynth in your blog. I am using ubuntu (karmic) only. I am now trying to install avisynth. Obviously there is no other way than compiling it.
cotcot, you won’t need to compile it, if I understand you correctly, you can download the AVISynth Windows executable installer and run it with Wine on Linux. I too am on Ubuntu Linux.
You’ll need Virtualdub too which is version 1.9.9 I think. For any mpeg2 files then VirtualdubMod is required although your’s are mpeg4 as you’re already aware.