Blurred Image Recognition

Abstract

Blurred image recognition architecture is divided into two steps. First, to deblur the image to find the underlying image. This process is known as image restoration. Restoration of the features is the most crucial task for image recognition. In other words, we can say that image restoration is a preprocessing required for the purpose of image recognition. Deblurring is an image restoration method of removing corruption like blurring artifacts from images. We have used blind deblurring i.e., we don’t have prior knowledge of point spread function (psf). Images can be damaged by blur, such as motion blur or blur associated with an out-of-focus lens. In this dissertation, we explain how an inertial measurement unit (IMU) can be used to generate blur-field or psf, from which we can get better image deblurring. We compare other method techniques from the proposed one and find out that the gyroscope-based method is superior than the rest. Deblurred image will have certain number of features that were present in the sharp image. We applied image recognition model on the deblurred image for the purpose of blurred image classification. We have compared different algorithm for image recognition that already perform good on sharp image. Deblurring and recognition algorithms are combined to create a novel architecture which can be used in many applications.