At present, the development of algorithms is particularly relevant high-level analysis of video sequences, including tracking of objects in a video stream and their implementation in information and communication systems. In the proposed approach the pretreatment step uses optical flow calculation for each frame video sequence, as well as low-level features (such as a representation of each pixel in the color space YIU). For moving objects optical flow is more important than for a stationary background and allows you to find the foreground objects in video sequences and adjust the weighting function in determining the histogram similar to the adjacent frames, taking into account the motion of objects.
Keywords: computer vision, tracker, tracking of object, optical flow, motion tracking
The paper presents a method of detecting distortions in the archival photo documents by texture analysis based on local binary patternss. The proposed approach consists of the following stages: pre-treatment and detection of scratches. Pre-treatment is used to reduce the effects of noise and small defects in the image. To recognize the class of defects using a support vector machine. The effectiveness of the new approach on several examples in the detection of defects.
Keywords: digital image processing, detection of defects, archival photographs, local binary patterns