Blur assessment using edge information

Quality of digital images is often impaired by blur artifacts in situation such as compression, focus error, relative motion and multimedia transmission. The loss of high frequency content leads to blurring effect in the image. In this paper, objective full-reference and no-reference blur assessments are presented to measure the degree of Gaussian blur in the image by using edge information. The degree of Gaussian blur of an image is the average of total edge widths over all detected edges. The performance of the blur assessment is also validated with subjective results. The results show that the blur assessment correlates relatively well with human perception. The blur measurement is applicable to numerous applications such as blur estimation in digital photography, image processing, printing or as a simple metric in comparing two images. The source code of this low computational complexity blur assessment is written in MATLAB program