You could create an AOI of the cloudy pixels and use this to extract by mask raster values in your non-cloudy scene and then merge the two.  You may have to use the target calculator to sum values and then "decode" these and reclassify. There may also be a way to reclassify those cloudy pixels values by your cloud free images. 

Dear Tamara,

Please tell me how i make AOI of cloudy pixels and where is target calculator is present

I do recommend to use fmask software for landsat 8. It will give you a cloud mask for clouds, cloud shadow, snow ..etc. The new file can be used with conditional statement that states if the file (cloud mask) has a certain value then obtain the pixel value from the other file (landsat 7). If you are using ERDAS, I can send you the model written in model maker. I am attaching a file for processing Landsat 8, you can look at the cloud mask part. The full manual is at Research Gate

example using ERDAS model maker

Thank You so much Sir Jawad T Al-Bakri.It is really helpful for me

Hello Usman. You can use the methodology to extract the cloud pixel values and after do the composition. Be aware to use the same or similar methodology to extract clouds in the Landsat 7 image in order to carry only pixels without clouds.

You can use several images, the concept is similar to the used in MVC (Holben, 1986). http://www.tandfonline.com/doi/pdf/10.1080/01431168608948945

Thank you sir

Consider image compositing. See David Roy for a key paper, linked below, with many others from his RG publications. Others to search on RG include Patrick Griffiths, Joanne White. 

Article Web-enabled Landsat Data (WELD): Landsat ETM+ composited mos...

Thanks Michael

Dear Jawad,

Your answer was quite helpful. But is this method suitable for cirrus clouds too??