Contents || Index
03/31/2009

Quality Tool

 

An experimental tool that assigns a quality factor to convergent beam diffraction images.  It 1) uses the Linear Hough Transform to find the lines (transforms them into dots), and 2) measures the ratio of line intensity to background intensity.

Short Instructions

Description of Commands

Description of Diagnostics

 


Measurement Commands

Hough Ratio Image            Hough transform of image divided by Hough transform of heavily filtered image (mean filter with flatten kernel radius).  The ratio is clipped to zero for denominator values less than 25% of the max Hough transform value of the smoothed image (denominator value).  This clipping gets rid of spurious signal of the ratio of small Hough values, which correspond to lines near the edge and corners of the image.

Q Measurement        (Quality Measurement).  I've intended these measurements for use on the Hough Ratio Image, but they will work on any front image.

Image Quality is a nebulous and flexible concept.  This button is the experimental part of the tool.  The measurements done by this button measure the contrast of the image - either by its range neglecting zero, or by its variance.   The EBSD images with more clear lines of greater contrast are images of higher quality.  The contrast of the lines appears in the Hough image as the brightness of the corresponding spots vs. the brightness of the background.  The background here is the image area near the spots, and not the areas at the top or bottom which are zero.  The zero valued areas correspond to lines outside the dimensions of the EBSD image.  The useful part of the Hough Ratio Image excludes these margins of zero value.

The outputs are text, tab delimited for direct cutting and pasting into Excel, as well as into any text editor for saving to a file.

Range...      Ignores zero value pixels in the Hough image.  Gives the minimum value excluding zero, maximum value, and difference between the two as the range.  The Hough Ratio Image may have borders with zero value, but this measurement ignores them, which makes manual selection of a rectangle not as necessary as with the Variance measurement.

Variance ...              Gives image  variance, min, max. The variance is the Square of pixel sample standard deviation.  Does not ignore zero value pixels in the Hough image, so specifying the rectangle to exclude the zero (black) margins gives a more meaningful quality value.


Diagnostic Commands

 

Flatten               Flattens image (a step in making the Hough Ratio Image).  This is a large scale unsharp mask.  A background image, generated from the original by mean filtering with a large kernel (Flatten Kernel Radius), is subtracted from the original.   This button enables you to examine the flattened image.

Bkg                    Displays the background image. This is for examination of the background image, a step in the Flatten calculation.

Flatten kernel radius   Smaller kernel radius reduces shading across image.  Shading  adds to the variance.

Filter kernel radiusMean filter (smoothing) of the result. Smoothes out pixel noise.  Noise also adds to the variance. Use Filter radius of zero for no filtering. 

 


Very Short Instructions

Open an EBSD image.

 

Make ratio Hough image from either the whole image or a rectangle within it, with 1) Hough Ratio Image.

 

Take range measurement on this image or on a rectangle within it using 2) Q Measurement using  one of the Range or Variance options.