Known errors:

**gsmooth (Nov 6, 1996):**"effect is to remove low spatial frequency components from an image" -> "effect is to remove high spatial frequency components from an image".**pixlog (Nov 6, 1996):**"Mapping function for exponential transform" -> "Mapping function for logarithmic transform" caption on Figure 1 (needs gif image hacking to fix).**affine (April 2, 1999):**We did not specify that positive angles meant clockwise rotation and that we were using image coordinates with +Y being down.**rotate (April 2, 1999):**We did not specify that positive angles meant clockwise rotation and that we were using image coordinates with +Y being down.Under these conditions, we get:

x = (x_1 - x_o) cos (theta) - (y_1 - y_0) sin (theta) + x_o

y = (x_1 - x_o) sin (theta) + (y_1 - y_0) cos (theta) + y_o

**adaptive thresholding operator (Aug 27, 1999):**The equation for the mean threshold should be .**exponential operator (July 10, 2000):**The equation in the caption in Figure 1 should be .**line detector operator (September 22, 2000):**The equation for testing the maximality of the convolution output should be .**Glossary entry for Pixel Connectivity (October 12, 2000):**The equation for defining the 4-neighborhood should be .**Weights for gaussian smoothing (September 20, 2001):**The weights should have been calculated by integrating over the pixel. It was unclear how they were calculated. New weights have been calculated, using a new sigma=1.0. New masks

and

are shown.**Weights for Laplacian of Gaussian mask (November 1, 2001):**The weights should have been calculated by integrating over the pixel. It was unclear how they were calculated. New weights have been calculated, using a new sigma=1.4. Some value have been rounded the wrong way to preserve the zero sum property. The new mask is:

**JAVA Operator Interative Exploration Under Internet Explorer (March 19, 2002):**All operator interative exploration now works; however the FFT and Histogram Equalization displays have layout problems that need fixing in the future.**JAVA Operator Tableau median smoothing operator (July 2, 2005):**Median filter in the tableau was fixed. Previously not correctly getting kernel values.**JAVA Connected components operator (Oct 4, 2005):**Fixed to work correctly with more than 256 components.**Laplacian of Gaussian small mask removed (Feb 15, 2006):**The 3rd small LoG mask -1,2,-1/2,-4,2/-1,2,-1 was removed as it is not a small mask approximation to the LoG.**Laplacian of Gaussian and Gaussian Smoothing JAVA code (March 7, 2006):**Replaced for loops with this structure:

`for(double ySubPixel = y - 0.5; ySubPixel < y + 0.6; ySubPixel +=0.1)`

by this structure:

`for(double ySubPixel = y - 0.5; ySubPixel < y + 0.55; ySubPixel +=0.1)`

because some users get asymmetric masks.**Exponential/`Raise to Power' Operator (November 1, 2006):**"As we can see from the formula, Q=0 yields P=c." should be "As we can see from the formula, P=0 yields Q=c.".**Logarithm Operator (November 1, 2006):**"Hence, both are suitable to enhance details contained in the high values." should be "Hence, both are suitable to enhance details contained in the low values.".**Connected Components Labeling (Nov 15, 2006):**"else if one or more of the neighbors have V={1}, assign one of the labels to p and make a note of the equivalences." -> "else if more than one of the neighbors have V={1}, assign one of the labels to p and make a note of the equivalences."**Fourier Transform (March 7, 2007):**In the "Guidelines for Use" section the example with the vertical stripes the f_{max} should be 1/(2 pixels) because the period is really 2 pixels. Similarly, the next equation should be 1/(4 pixels).**Fourier Transform (Aug 23, 2007):**In the "How It Works" section, the use of (i,j) for image coordinates is replaced by (a,b) because of the potential for confusion with i or j as the square root of -1.**Gaussian Smoothing (April 4, 2008):**"mean filter with width 7" -> "mean filter with width 5".**Conservative Smoothing (July 21, 2008):**conservative smoothing interactive demo fixed (missing executable in jar make).**Roberts Cross (September 14, 2009):**Edge orientation estimate from arctan()-3/4 pi to arctan()+1/4 pi.**Fourier Transform (December 29, 2009):**Clarity improved in several places by making explicit where the magnitude of the Fourier transiform is displayed.**Fourier Transform (February 4, 2011):**Correcting 1/(N*N) normalization to appear on only inverse transform, and adding a note about about equivalence.**Gaussian Smoothing (June 11, 2013):**Explaining in more detail how the example smoothing mask values were calculated.**Fourier Transform JAVA demo (March 16, 2015):**Added a note about about ensuring that the images are square, otherwise the FFT may not work.**Hit-and-miss operator (April 21, 2016):**Added mask 3c for completeness.