Evaluation and comparison of techniques for the reconstruction of the point spread function of uniform linear motion blur images
In the digital image processing area, it is common to find different types of degradation such as motion blur, which is caused by the relative movement between camera and observed object. This produces a low-contrast trace on the image that follows the object’s displacement. If the relative velocity is constant and the blur is invariant along the entire image, the resulting blur can be modeled by means of the Point Spread Function (PSF) using the trace’s length and the angle parameters. It was evaluated into this research, the accuracy in the estimation of the angle and the length parameters, and the robustness to the Additive White Gaussian Noise of a set of spatial and frequency approaches for the reconstruction of the PSF. It is important to highlight that the algorithms’ processing time was considered. There were used 20 synthetically degraded images at a resolution of 512x512 pixels. There were evaluated as well, five of the best-known techniques for estimating the angle and three for the length of the PSF. The experimental results revealed that the techniques with the lowest absolute mean error for the estimation of the angle and the length of the PSF in noise-free images are the 2D Cepstrum Transform and the 1D Cepstrum Transforms, respectively.
Full Text:PDF (Español)
- There are currently no refbacks.