Video sequences contains multiple frames therefore their quality is estimated by determining individual quality metric of each frame then apply the temporal masking affect. However, the integration of each frame’s quality metric into one score is very important because each video frame has different spatial features hence have different quality metric. There are several methods available to combine the metric into one score like averaging, linear weighting, worst frames averaging etc. Taking the average of each frame’s score is not very useful as humans give more attention to the worst values (most distorted frame) while rating their values. In this paper we evaluated the performance of different integration methods and a different approach is proposed which includes the average of worst selected frames which is discussed in later sections. The work is tested on LIVE video database which consists of 40 video sequences. They have provided the mean opinion scores for each video with the database. The correlation coefficient of 88.21% is achieved when tested with the best model designed.