Panoramic image stabilization algorithm based on region segmentation and fusion

Vehicle infrared panoramic scanning imaging system has the characteristics of column single imaging and 360 full field view, which result in that traditional electronic image stabilization methods could not be directly applied, therefore, the panoramic image stabilization algorithm based on region segmentation and fusion was proposed. First, the image pre-compensation was completed by local column offset adjustment method. Then the panoramic image was divided by the area, i.e., the front region, the right region, the end region and the left region, based on vehicle forward direction. Then each region used different image stabilization models for stabilization according to imaging characteristics, and the sliding window strategy was used to shorten the computation time in motion estimation process, the undefined area reconstruction method was used to compensate for missing information in motion compensation process as well. Finally, the local area expansion strategy and the progressive fade-weighted average fusion strategy were adopted to complete the region's splicing and fusion, resulting in the seamless splicing effect of panoramic images. Experimental results indicate that the proposed method which can meet the requirements of engineering applications can effectively solve image stabilization problems of the system in the course of the vehicle moving forward, the precision of Peak Signal to Noise Ratio(PSNR) is increased by 14.7% compared with the original image sequence, running time is reduced to the 1/10 of traditional algorithms.