Evaluation of Scumming Printing Defect by Using Computer Vision-based Bit Plane Slicing Method

In offset lithography scumming print defect occurs when non-image areas of the image carrier accepts ink and transfer it to the blanket. The quality of printed image is compromised with the presence of scumming pixels. Visual impact in terms of readability affected due to this unwanted print defect. Moreover, archival process through scanning and OCR can be interfered by the presence of scumming. Various press conditions are responsible for this phenomenon. Generally, this kind of printing problem is detected manually and preventive measure is taken according to the factor responsible for the occurrence of this defect. Traditionally, detection of this defect relies heavily on manual inspection, which is prone to inconsistency and human error. In response to this challenge, the present study proposes an automated, low-cost computer vision-based approach for detecting scumming defects using bit-plane slicing and discrete cosine transform (DCT ). In this present study a computer vision-based approach has been proposed to evaluate this kind of print defects and its percentage of occurrence over the existing detection procedure. As the density difference of scum pixels and print pixels are not that much differentiable, it becomes a challenging task to identify the scum pixels from print pixels in the present computer vision-based study. Intensity of pixels can also be varied in different kind of images which makes the segregation of scum pixels even more difficult. An automated algorithm selects the bit plane containing the most scum-relevant information, followed by DCT filtering and two-stage segmentation to isolate print and scum pixels. To segment the scum pixels from the print pixels, an adaptive thresholding method with a proposed range of intensity values have been applied to high key, low key and mid key images. The proposed method demonstrated high agreement with manual visual assessment, achieving consistent results across high-key, mid-key, and low-key images. Quantitative scumming percentages are calculated, offering a reproducible metric for defect severity. The result shows the efficiency of the proposed algorithm as compared to present subjective method.