This paper aims at implementing an integrated super resolution reconstruction algorithm to interpolate the missing pixels in the grid to create a high resolution image for a specific purpose. One of the most important application areas of super resolution reconstruction is video surveillance for the purpose of public security. Although, the video surveillance technology are going under tremendous transformation from analog generation to IP-based systems. However, their replacement rate is still not encouraging due to installation and operating costs as well as low output video quality. To overcome this problem, we propose a new hybrid model to integrate super resolution reconstruction into video surveillance. Our proposed algorithm is based on interpolation of cropped low resolution frames extracted from a low quality video surveillance sequence for effective and efficient reconstruction of a high resolution license plate recognition image. Our super resolved image utilizing multiple frames provides far more detail information than any interpolated image from a single frame. The proposed algorithm requires a relatively small number of self extracted low resolution frames from a low quality input sequence. This is important for practical applications, because if a large number of low resolution frames were required the accumulation of imaging errors would adversely affect the reconstruction accuracy. We apply our proposed algorithm to a real sequence from video surveillance and compare our results with those obtained via well-established techniques. Experimental results show that the proposed algorithm performs much better than the conventional MISO super resolution techniques. We also noticed significant reduction in the computational cost and memory requirement during the whole reconstruction process.
展开▼
