Non–referenced quality assessment of image processing methods in infrared non-destructive testing based on higher order statistics

摘要

Current image processing techniques require to collect a large amount of data to enhance the defect's visibility in materials, leading to the best-quality image must be found exhaustively from the whole sequence. In this work, we study the adequate implementation of Infrared Thermography based on Higher Order Statistics (IRTHOS) technique for Infrared Non- Destructive Testing (IRNDT) where a defect can be detected from a single enhanced image, avoiding the use of the whole image data to achieve the best-quality image. For validation purposes, we compare the performance of IRTHOS among the common techniques used for IRNDT. Comparison is carried out by quality assessment of processed images of considered techniques. We use a Non-Referenced (NR) measure for Image Quality Assessment (IQA), giving as a result that IRTHOS achieves a 4.68% higher quality for TSR first derivative, than the best-quality image found. However, the image processed by the considered technique exhibits singularities due to net structure and geometry of the material.