Creation of image models for inspecting visual flaws on capacitive touch screens

Touch screens (TSs) are commonly applied in many electronic appliances such as smartphones, tablets, etc. Currently, capacitive touch screens (CTSs) are the main touch technology of screen panels due to many excellent electronic properties. Problems exist in inspecting flaws inlaid in appearances of CTSs with structural patterns. Area flaws are a type of common visual defect that comprises dust, bubbles, ripple marks, and other flaws of bigger sizes. These flaws have the attributes of low contrast, brightness with slow changes, unusual and non-orientation forms, and sometimes both bright and dark flaws existing at the same time in a region. This paper suggests image models based on transformation filtering to inspect the area flaws on appearances of CTSs. We apply the Haar wavelet transform with flat zone filtering technique to eliminate the structural patterns of background by means of filtering an approximate sub-image of a breakdown wavelet domain image. Subsequently, the filtered image is reversely transformed to obtain a rebuilt image in spatial domain. Last, the rebuilt image with intensified flaws can be simply partitioned into three species (black flaws, gray flaws, and white background) by using a statistical interval estimation method. Therefore, the intricate area flaws are precisely identified by the suggested scheme. We contrast our approach with three traditional methods with real samples under complex background and conduct quantitative comparisons. The effectiveness and accuracy of the developed image models are confirmed by expert assessments, as well as by comparative analysis with the known methods in the field of spatial localizations and production-related effects of flaw detection.