Abstract

The image classification rate of a bad quality image can be improved by performing\npre- processing techniques on it. Pre-processing techniques like Mean filter, Image\nnormalization, DeCarrstrech, luminizer, linear-contrast Adjustment, Blur removal can help in\nremoving noise, uneven color intensity distribution, improving the luminance of low light images,\nunintentional blur detection, and removal, etc. While implementing pre-processing techniques it\nis necessary to identify which techniques are to be used and also the order of using them to get an\noptimized result. While the existing system was developed to handle a singular problem, our\nproposed system is an ensembled system to handle multiple image defects. To automate this\npipeline, we use Neural Image Assessment (NIMA) which will identify the defects in the image\nand the designed system will choose the appropriate image preprocessing channel which would\nyield the best results for image manipulations like feature extraction, image classification, object\ndetection, facial recognition, and identification, etc.