Abstract: Serial time-stretch imaging technique with its high line-scan rate can observe and record fast non-repetitive events. However, it is inevitable that a fast imaging system would generate a massive amount of data. This paper proposes a data compression methodbased on differential detection and run-length encoding for time-stretch imaging technique, which can efficiently solve the existing problem of data volume in the back-end digital signal processing. Differential detection can eliminate identical signals and only distinguish the difference between consecutive scans, while the run-length encoding algorithm is suitable for encoding the same repeated signals, to further improve the effectiveness of run-length encoding algorithm. In the experiment, a 77.76 MHz line-scan imaging system is demonstrated with the resolution test target, red blood cell and MCF cell. It turns out that the compression rate of 8.47 is proved. Through analysis, it is found that the method with differential detection can achieve higher compression rate than the method without differential detection, especially when the sampling resolution is low. The structured similarity index measurement(SSIM) calculation between the original image and the reconstructed image shows high quality images can be reconstructed after compression by the proposed method.