Introduction to Robust Real-Time Super-Resolution on FPGA:
The imaging system is generally based on a sensor known as image sensor which is a 2D array of pixels and enables to covert incident light to the electrical signal array. there are two kinds of resolutions that determine information quality collected by the sensor and it includes the temporal and spatial resolution. The spatial density of the photo dies and induced blur is responsible for the spatial resolution. An intuitive solution that is responsible for the spatial resolution enhancement is reducing the size of pixels thus increasing the density of the pixels. The smaller the photo dies the small will be the incident light amount. In order to get the desired signal it is important to make a longer exposure.
In absence of relative motion between scene and camera the light amount reduction can be easily determined by enhancing the pixel time of exposure which means increasing the time for integration of the photodiodes. In real either the camera shakes while taking the shot or the object is shaking during the period of integration. More number of my integration time spans require for real world frame. The result thus suffers from the motion blur reducing the temporal resolution.
In this thesis, an FPGA based motion deblurring system is proposed that make use of the image sensor configured with large pixel areas on the regions of motion. In order to compensate for the low spatial resolution FPGA based SR is used. The FPGA is implemented with the IBP algorithm and that too after modifying the account for the LR simple addictive noses. Various noise levels are evaluated for this system keeping in mind the several parameters like the initial kind of approximation.
LR sample numbers and the length of the words. According to the results which gives noise static informations in the algorithm enhance the properties of the process of interative dramatically, in the noisy environment. This also leads to a more profound scheme.