In this work a machine vision system for inspecting pharmaceutical color tablets is presented and implemented. Nonparametric clustering based segmentation is faster and thus more appropriate for real-time applications. Two nonparametric clustering based methods, Nearest Neighbor algorithm and MaxShift algorithm are worked in RGB and HSV color spaces as the segmentation step. The implemented algorithm allows the system to detect the missing and broken tablets, tablet fragments, and the color, size, and shape of individual tablets in pharmaceutical blisters, in real-time. System has two operation modes called &amp / #8216 / &amp / #8216 / training&amp / #8217 / &amp / #8217 / and &amp / #8216 / &amp / #8216 / inspection&amp / #8217 / &amp / #8217 / mode, respectively. Operator selects one point on any tablet in a defect-free training captured image in the &amp / #8216 / &amp / #8216 / training&amp / #8217 / &amp / #8217 / mode. In the correction step an optimization algorithm is required, for which Powell and Downhill Simplex methods are used. Captured image is then corrected for spatial color nonuniformity, segmented, and the position, size, shape, and color of each tablet are extracted in the training mode. The correction and segmentation models / the extracted features generated in the training mode is saved with the user defined values to form the model. Each acquired image in the inspection mode is corrected and segmented according to the blister model and then the blisters are classified as &amp / #8216 / &amp / #8216 / good&amp / #8217 / &amp / #8217 / or &amp / #8216 / &amp / #8216 / bad&amp / #8217 / &amp / #8217 / by comparing the extracted feature values with the user defined tolerances stored in the blister model.

<p>This thesis describes and evaluates a number of algorithms for reducing fixed pattern noise in image sequences. Fixed pattern noise is the dominantnoise component for many infrared detector systems, perceived as a superimposed pattern that is approximately constant for all image frames. </p><p>Primarily, methods based on estimation of the movement between individual image frames are studied. Using scene-matching techniques, global motion between frames can be successfully registered with sub-pixel accuracy. This allows each scene pixel to be traced along a path of individual detector elements. Assuming a static scene, differences in pixel intensities are caused by fixed pattern noise that can be estimated and removed. </p><p>The algorithms have been tested by using real image data from existing infrared imaging systems with good results. The tests include both a two-dimensional focal plane array detector and a linear scanning one-dimensional detector, in different scene conditions.</p>

This thesis describes and evaluates a number of algorithms for reducing fixed pattern noise in image sequences. Fixed pattern noise is the dominantnoise component for many infrared detector systems, perceived as a superimposed pattern that is approximately constant for all image frames. Primarily, methods based on estimation of the movement between individual image frames are studied. Using scene-matching techniques, global motion between frames can be successfully registered with sub-pixel accuracy. This allows each scene pixel to be traced along a path of individual detector elements. Assuming a static scene, differences in pixel intensities are caused by fixed pattern noise that can be estimated and removed. The algorithms have been tested by using real image data from existing infrared imaging systems with good results. The tests include both a two-dimensional focal plane array detector and a linear scanning one-dimensional detector, in different scene conditions.

No description available.

No description available.