newsletter archives - October, 2003

Volume 8. Issue 11.

[page 1][page 2][page 3]

Document Imaging Effect and Filter Functions What is image filtering? Filtering of an image means replacing the pixel values with new ones computed from only the original pixel value, the original pixel value and the local neighboring pixels, or the original pixel value and the global image properties such as histogram. The Document Imaging SDK presents many filter features for 1, 8 and even 24 bit per pixel grayscale and color images compartmentalized to filtering groups. What is image filtering good for? The three typical goals of image filtering are: · Quality improvement such as correction of noise, corrupted, low or high exposed, images, etc. · Retrieving special information (e.g. geometrical features) of images such as edge detection and binary hit-miss filters. · Simulation of external effects such as frequency high and low pass filters and motion blur. Effect and Filter Groups in Document Imaging · Adaptive Filters: These filters are typically used for noise removal. The adaptivity of these filters is based on the fact that the filter is not using just one or more exact parameters but instead some retrieved information of the parameter specified pixels are also tuning the filtering algorithm. There are 2 adaptive filters available and they are Adaptive-DW-MTM and Adaptive MMSE filters. · Artistic Filters: The artistic filters don’t have an exact goal. These filters have different properties, can be used for anything from filtering on histogram to generating motion blur on images. There are 8 artistic filters available and they include Auto Contrast, Motion Blur, Posterizing, and more. · Edge Detector Filters: These filters are used for retrieving edge information from the input images. Three complex edge detectors are presented, all of them can be used to detect the edges of different objects from geometrical figures to human faces. There are 3 edge-detector filters available and they are Marr-Hildreth, Canny and Shen-Castan edge detectors. · Morphological Filters: Morphological filters are usually used to filter binary images (some cases grayscale 8 bit per pixel images). Morphological filtering are typically used to remove noise or improve quality of scanned fax pages, writings, drawings, etc. There are 10 morphological filters available and they include Dilation, Erosion, Closing, Opening, Hit-Miss, Skeletonizing, Top-Hat, and more. · Nonlinear Filters: These filters are typically used to remove several types of noise corrupting the input image. Due to their nonlinearity, these filters are usually slower than the other filters, but a well chosen and well parametered nonlinear filter can achieve very good results in noise removing. There are 11 nonlinear filters available which include several types of Mean filters, Median filter, Weighted Median filter, and more. · Spatial Filters: This filter group contains the linear filters which can be used to remove noise, to enhance or smooth edges, or to find edges on the image. There are 9 spatial filters available and they include Low and High pass filters, Gradient, Laplace, Uniform filters, and more. · Spatial Frequency Filters: This is a very special filter group which contains the frequency filter functions. Frequency filters are filtering in frequency domain, first they compute the spectra of the image, then they filter the spectra, and finally compute the resulting image from the filtered spectra. There are 7 spatial frequency filters available and they are Low and High pass filters, Enhance, Homomorphic, Inverse, Wiener filters and Fast Fourier transformation. The filters of the Document Imaging SDK/ActiveX are implemented in one dynamic library (bifilter.dll). All of the filters can be used by giving the parameters of the filtering, and all of them can show a preview dialog. It’s important to note that if a filter uses information of the whole image (global filtering), it is possible that the result of filtering will be different from the filtered cropped image of the preview dialog.

[page 3]