Image Processing Toolbox Release Notes Previous page   Next Page

Major Bug Fixes

This section describes the bug fixes in these Image Processing Toolbox releases.

Bug Fixes in Version 5

The following are important bug fixes in Version 5 of the Image Processing Toolbox.

Canny Edge Detector Handles Constant-Valued (Flat) Images

The Canny edge detector now accepts single-valued images, also called monotone images, constant-value images, or flat images. Instead of issuing an error when an input image is single-valued, the edge function used with Canny edge detection now returns an output image containing all zeros, indicating that it found no edges.

imcomplement Returns Correct Answer for Signed Integer Input

imcomplement was incorrectly calculating the complement of an image with signed integer data type. This problem has been fixed.

imlincomb Correctly Handles int16 Data Combined with Scalar Having a 0.5 Fractional Part

imlincomb was giving an incorrect answer when combining int16 data with a scalar that had 0.5 as a fractional part. This problem has been fixed.

iradon Introduced a Vertical Shift of One Pixel

The iradon function now correctly calculates the vertical origin of the input projections. Previously the calculated origin was off by one for inputs with an even number of projection samples. The effect of this problem could be observed by computing the Radon transform (using radon) of a test pattern containing horizontal edges, followed by computing the inverse Radon transform (using iradon). Careful comparison of the test pattern with the output of iradon showed a vertical misregistration of one pixel.

imshow Correctly Renders Indexed Images with Colormaps Having More Than 256 Colors

On the Windows platform, imshow now sets the figure's 'Renderer' property to 'zbuffer' for indexed images with associated colormaps having more than 256 entries, so they render correctly. Previously these images would appear black.

Bug Fixes in Version 5.0.1

The Image Processing Toolbox 5.0.1 includes the following bug fixes.

The DCT Image Compression Demo Now Calculates Mean Squared Error Correctly

The 2-D DCT Image Compression demo (dctdemo) previously calculated the mean squared error incorrectly.

Choose Colormap Tool Now Works with Compiled Applications

The Choose Colormap tool now saves the colormap selection when running in a compiled version of the Image Tool.

Image in Pixel Region Tool Now Updates When Using the Adjust Contrast Tool

Previously, if you used imtool to display a grayscale image, launched the Pixel Region tool, and then used the Adjust Contrast tool to adjust the contrast of the image, the pixel colors in the Pixel Region tool did not update properly. This has been fixed.

Bug Fixes in Version 5.0.2

The Image Processing Toolbox 5.0.2 includes the following bug fixes.

Major Revisions to Fan-Beam Functions

This release includes numerous updates and improvements to the fan-beam functions: fanbeam, ifanbeam, fan2para, and para2fan. The fixes include improved calculations, improved documentation, and examples.

For example, fanbeam now returns the correct sensor locations when the geometry is 'line'. The ifanbeam and fan2para now consistently use the correct default value for the 'FanSensorSpacing' parameter. If you tried the fan-beam functions in a previous release, you might try them again to take advantage of these improvements.

In addition to the functional changes, many improvements to the documentation of the fan-beam functions have been made.

fanbeam help now includes

The help for the ifanbeam function now includes an example that shows how to use the 'minimal' coverage parameter.

Changes to the DICOM Functions

The following fixes have been made to the dicomread and dicomwrite functions.

Function
Bug Fixes
dicomread
No longer errors when reading files that contain extraneous pixel data; instead, dicomread issues a warning message. However, if the file does not contain enough pixel data, dicomread issues an error.
dicomwrite
  • No longer is case sensitive when parsing input parameters. For example, you can specify either 'CreateMode' or 'createmode'.
  • Preserves the full precision of data converted to decimal string metadata. Previously, dicomwrite limited precision to six digits.
  • No longer errors when writing files with metadata values that must be stored as a decimal string or integer string. Now, when writing private data attributes (attributes that are not listed in the DICOM data dictionary), dicomwrite assigns the attributes the type UN (for unknown) and writes the data to the file as a byte-for-byte copy of its in-memory representation. Because dicomwrite writes the file with explicit value representation (VR), the file might have a different VR value, but the data will be the same.
  • Includes the TriggerTime field for additional values of ScanOptions, including 'CT'. Previously, dicomwrite only included the TriggerTime attribute if the ScanOptions field indicated a gated heart MR.
  • No longer issues an Unsupported SOP class error message if, when 'create' mode is specified, semantic verification is not available for an information object. Instead, dicomwrite issues a more helpful message indicating that it might be able to write the data if the mode was 'copy', rather than 'create'. In 'copy' mode, dicomwrite only performs syntactic checking, not semantic verification. Consequently, dicomwrite can write many more types of DICOM files in 'copy' mode than it can in 'create' mode. See the dicomwrite reference page for important information about data integrity.

Changes to Image Tool and Modular Interactive Tools

The following fixes have been made to the Image Tool and other modular interactive tools.

Changes to the imshow Function

Fixes to Other Functions

The following tables lists fixes that have been done to other toolbox functions.

Function
Enhancement
applycform
Now correctly handles profiles that contain a gamut tag.
cpcorr

Now is more numerically robust. For this release, the subfunction findpeak, which cpcorr calls, has been improved and is now a private function, rather than a subfunction.

imhist
No longer causes a docked figure window to become undocked.
imrotate
Now correctly rotates N-dimensional arrays, where N is greater than 3. In previous releases, imrotate would accept N-D arrays but only return a 3-D array.
normxcorr2

Now always returns real values. In previous releases, due to roundoff error, some sets of input data caused the normxcorr2 function to return a complex valued matrix of correlation coefficients.

pixval
Now works correctly with binary images.
rgb2ind

Now returns a correct output image when called with the syntax

  • rgb2ind(rgb,n,'nodither') 
    
where n is greater than 256.

Fixes to Image Processing Toolbox Deployment Issues


Previous page  Obsolete and Removed Functions Upgrading from an Earlier Release Next page

© 1994-2005 The MathWorks, Inc.