Can anyone help me regarding the what kind of image pre-processing techniques should be used for the landsat images downloaded from http://earthexplorer.usgs.gov/ to create land-use land-cover (LULC) map?
Atmospheric correction is required for all time series studies, as indicated in your comments. Dark object (aka Dark Body) subtraction is a common method to accomplish this as long as a suitable dark object, such as a water body, exists in the scenes.
@Mr. Che and @Aaron have given great answers regarding two different things to do, but i would like to point out that there is actually a quick and easy way to use the landsat data without dealing with implementing the fmask algorithm or doing the atmospheric correction yourself: by using the surface reflectance data processed based on Landsat data. They are available for download at EarthExplorer as well. These data have been processed so that there is no need for you to do either georegistration or atmospheric correction any more (most of the time). Plus together with the surface reflectance data they also provide various cloud masks, including fmask-based ones, so you don't need to implement the algorithms yourself.
I'd also like to answer your question by saying that there is no universal way to preprocess data for land cover and land use mapping. You always have to decide what is necessary and what is not based on your specific needs and the data you have access to. But surface reflectance data is usually a good starting point.
P.s how to remove cloud for landsat 7 images?
You can try to use automate standalone algorithm Fmask described here.
We're looking for long answers that provide some explanation and context. Don't just give a one-line answer; explain why your answer is right, ideally with citations. Answers that don't include explanations may be removed.