Authors: Santhosh SG, Sampath Kumar
Abstract: As digital data rapidly increases, there will be a corresponding depletion of textural data for various uses, as the number of image-based documents with usable text continues to rise. But too often, this is complicated by the obstacles of storing images with distortion, algorithmic font types, misaligned printed text, random text orientation and other forms of noise. For considering image-based documents with bilingual documents like government forms, educational transcripts, medical records, and business receipts with multiple integrated languages across a single document, become more complex and piled on these particular layers of challenges. The research paper "Optimizing Evaluation Metrics with PSNR, AMBE, and F1 Score to ensure Consistency in Document Improvement and Consistency in Classification Accuracy" is intended to examine some of the issues that both have been examined and expressed. This effort confronted the problems this work has discussed and expressed through consistent and reliable classification accuracy; AMBE determines the interference with the brightness distribution; and PSNR applies a clarity "score" to determine the clarity of an image. Combined, the three metrics present a possible framework to enhance the reliability and consistency of document processing system.
DOI:
