CORRELATION AND RELAXATION LABELING - AN EXPERIMENTAL INVESTIGATION ON FAST ALGORITHMS

This paper compares experimental results between three popular matchin g functions: the cross-correlation coefficient (CCC); the sum of squar ed difference (SSD); and sum of the absolute value of difference (SAVD ), within our newly developed correlation-relaxation (C-R) framework ( Wu 1995). The C-R framework is a general method for determining optica l flow and has been applied to determining cloud motion from satellite images. SSD and SAVD are simpler and faster functions to calculate, w hen compared with CCC, and their uses can lead to significant savings in computer time in the initial selection of displacement candidates. Given that the image distortion is Gaussian noise, and the motion is t ranslational, the study shows that while computationally more expensiv e, the performance of the CCC function is better, or at least no worse , than using SSD and SAVD in the selection of initial displacement can didates. Similarly, the performance of SSD is better, or no worse, tha n using SAVD. Computationally, SSD is the fastest among the three func tions. In the presence of high level distortion, however, the poor qua lity of initial candidates selected using SSD and SAVD usually means a large number of iterations of the subsequent relaxation labelling pro cess. In contrast, the CCC function gives high quality initial candida tes, and only a small number of iterations are needed. The CCC functio n also usually leads to better final quality in motion estimations tha n that produced using the SSD or the SAVD function in the C-R algorith m. In the presence of moderate and low level distortion, however, the performance of SSD can be adequate, and its use can lead to faster pro cessing without much sacrifice to the overall motion estimation qualit y.