Image matching using peak signal-to-noise ratio-based occlusion detection
A new identification mechanism is introduced for the purpose of locating objects partially occluded under low peak signal-to-noise ratio (PSNR) environment in two-dimensional grey scale image.The proposed occlusion detection is based on the utilisation of the fact that the higher the PSNR, the less the impairment of the image. Most existing methods require a training process before recognition tasks or fail to obtain good results when objects are partially occluded under low PSNR environment. The new identification mechanism uses a self-adaptively adjusted threshold for providing a more exact occlusion detection and a correlation-coefficient evaluation process for reducing false positives, allowing for better accuracy in classifying and locating partially occluded objects under low PSNR. The performance of the proposed framework is confirmed through 100 occluded images with various noise levels. The experimental results show that the new proposed algorithm is not only more robust against image noise and partial occlusion, but also provides a significantly improved object localisation/recognition performance when compared with normalised cross-correlation and selective correlation coefficient.