Application of Digital Images Blur Estimation Methods for Audiovisual Monitoring
Keywords:
Digital Image Processing, Image Quality Estimation, Image Blur, Face Segmentation and RecognitionAbstract
Purpose: Studying and developing mathematical and software tools to estimate blur and other quality metrics of digital images.
Results: The paper presents classification and analysis of modern blur estimation methods used for image processing in real-time
systems. In the experiment, for higher processing speed and more reliable evaluation of an image fragment quality, a number of
segmentation modes were applied: the entire image, a face extracted from the image and a 200x200 pixel area on the face. The proposed technique for extracting the face area from an image and estimating its blur on basis of statistic analysis of pixel brightness coefficients successfully recognized 94% of non-distorted images obtained by automatic registration of conference participants in an intelligent meeting room. Practical relevance: The proposed solutions serve to speed up the processes of automatic recognition of digital images and can be efficiently used in a wide range of computer vision applications.