Research On The Technology Of Electronic Image Stabilization Based On Fast Feature Matching

The camera system, such as Handheld photography equipment, On-board camerasystem,airplane,Ships photography platform and so on. Their carriers are shaking, so the video is notsteady. Video shaky can lead to investigators’miscalculation because of visual fatigue, and alsoincrease the difficulty of the subsequent image processing. Electronics image stabilization is a newmethod to stabilize the video.,which take use of the image processing technology. Because EIS hasthe advantages: High precision,small volume, light-weight,low power consumption and low cost,researchers have paid more and attention to it in the recent years.The method of this paper is based on feature matching,which can stabilize complex shakingvideo includes Translation and rotation.Global motion estimation is the most improtant step ofelectronic image stability, Improving the precision and speed of global motion estimation willmaximize the system performance. This paper proposes a the global motion estimationalgorithmbased on KLT-RANSAC.Firstly detecte the feature points, use the distance constraintmechanism to solove the problem of feature points crowding in local area. Which can eliminate thefeature points with smaller characteristic value, keep the feature points with bigger characteristicvalue.so that the extracted feature points distribute equably, and the quantity is moderate. Then FASTfeature point position precision is improved to the sub-pixel level according to the principle of vectororthogonality: the vector, which is from feature corner in sub-pixel level to the neighborhood point, isorthogonal to the gradient of this point. Feature points will be matched, the process is divided into twosteps: coarse matching and precise matching, in the first step, match the feature points fast with KLTalgorithm, RANSAC algorithm is adopted in the second step to eliminate the error matched pointswhich distribute on the moving objects, it avoids the interference of local movement on the imagestabilization. The global motion parameters can be more accurate by means of the correct precisematching result. As to motion compensation module, quadratic interpolation algorithm is adopted toget high quality image without jagged equity curve and ripple wave. Finally the stable video isobtained.Experimental Results show that, electronic image stabilization algorithm based onKLT-RANSAC global motion estimation can get global motion parameters with greater speed andaccuracy. The video frames look like the smooth nature, contain no jitter and distortion. The electronicimage stabilization system can satisfy the requirement of real-time,can stabilize the video with localmotion and shaking in any complex form.