Qualitative trend analysis (QTA) is a data driven approach and has been widely used in data compression, data monitoring, fault diagnosis and etc. However, for the traditional qualitative trend analysis, the qualitative trend is usually extracted by a fixed window and thresholds used are always dependent on human experience. So the qualitative trend extracted cannot represent the real trend reasonably. A novel multi-scale qualitative trend analysis algorithm is proposed to extract the qualitative trends. The process data is fitted linearly in a sliding window. The window can be extended or reduced to determine a segment. After all the segments for the data are determined, a qualitative trend of the data has been extracted. And then the initial width of the sliding window is changed to repeat above work in different scales. All the trends in different scales are ranked according to the fit index (F.I.) in decreasing order. The segments of the first one are sent to trend identification. The segments are indentified as “Increasing”, “Decreasing”,” Steady” and the qualitative trend is obtained. The result of case study proved that the trend can be extracted and identified efficiently with high accuracy using the algorithm.