Method for detecting optical defects in transparencies

A method of characterizing optical defects in a transparency, comprising the steps of:

• providing a digital image of the transparency, the digital image including a plurality of image pixels;
• identifying candidate pixels among the image pixels, the candidate pixels comprising at least one candidate defect;
• identifying the location of each one of the candidate pixels;
• clustering the candidate pixels into at least one pixel cluster based upon the relative locations of the candidate pixels;
• and
• comparing the quantity of the candidate pixels in the pixel cluster to a pixel quantity threshold to identify the candidate defect as at least one of an optical defect and an image defect.

A method of detecting optical defects in a transparency having a viewing portion, the method comprising the steps of:

• providing a digital image of the transparency, the digital image including a plurality of image pixels each having a grayscale intensity;
• detecting a perimeter of the viewing portion by performing at least one of the following:
• selecting a predefined perimeter of the viewing portion;
• comparing the change in grayscale intensity across a series of the pixels to a threshold intensity change rate;
• detecting candidate defects in the viewing portion by performing the following:
• calculating an intensity gradient across each one of adjacent pairs of the image pixels, the intensity gradient representing the difference in the grayscale intensities of the adjacent pairs of the image pixels;
• assigning to each image pixel a gradient value comprising a maximum of the absolute value of the intensity gradients associated with the image pixel;
• constructing a gradient image comprising the gradient values assigned to corresponding ones of the image pixels;
• identifying as candidate pixels the image pixels having a gradient value exceeding a gradient threshold, the candidate pixels comprising the candidate defect.
• classifying the candidate defects by performing the following:
• identifying the location of each one of the candidate pixels;
• clustering the candidate pixels into at least one pixel cluster based upon the locations of the candidate pixel;
• comparing the quantity of the candidate pixels in the pixel cluster to a pixel quantity threshold to identify the candidate defect as at least one of an optical defect and an image defect;
• determining a boundary of the optical defect by using an energy function;
• and
• characterizing the optical defects in at least one of location, size and shape.