Real-time Restoration Of Motion Blurred Image Based On GPU-Accelerated

Motion-blurring is caused by the relative motion between the camera and the object. It is a ubiquitous matter. Such as the photo shot at the plane or space aerocraft, the photo of vehicle moving at high speed, missile flying at the battlefield. The restoration of images blurred by motion is one of the main questions for discussion. It has realistic significance, applying in many fields. Such as military affairs, traffic, medical images, and detective field.Nevertheless, image restoration need process a mass of data, and restoration algorithm is complex, so it will influence the speed of process.With constant improvement of Video Monitoring System demands and the development of image restoration algorithms, people put forward two requirements of "a lot of information" and "fast" for restoration algorithm base on ensuring image restoration quality.At present,the two requirements are mainly implementing on 25 frames /s real-time restoration of 512×512 video blurred image. But the methods of meeting those requirements are mainly using the scheme of multiprocessor, in another word, that is increasing the speed of image restoration on the premise of increasing parallel hardware equipment, but the experimental environment is uneasily frame, and it costs relatively large the cost of hardware.Graphics processing unit (GPU) has been developing rapidly in recent years, and as a result, various applications associated with computer graphics advance greatly. At the same time, the highly processing power, parallelism and programmability available nowadays on the contemporary GPU provide an ideal platform on which the general-purpose such as digital image processing computation could be made. Aiming at the problem of relatively large the cost of hardware, this paper raised A novel method of real-time restoration of motion blurred image based on GPU-accelerated, and map winner filtering algorithm to texture-rendering on consumer-level graphics cards, the algorithm is accelerated through GPU by taking the advantage of parallel computation and fast floating-point computation. This method speeds up the execution of wiener filtering and realizes real-time restoration of 512×512 image in ordinary PC. The method meets real-time the requirement and provide a feasible approach to real-time restoration of motion blurred image for the common customers.