under the supervision of Noël BONNET

Description of plug-ins |
Download plug-ins |

Performs image smoothing with a limited degradation of contrast (smoothing is not performed across edges)

Two versions are available (in two different plugins):

the version originating from Perona and Malik (1990) [MORE ...]

the version originating from Black et al. (1998) [MORE ...]

The aim of this plug-in is to correct shading artefacts a posteriori, i.e. without any additionally recorded image.

2 modes are available: automatic and semi-automatic.

In the automatic mode, a number (chosen by the user) of markers are spread regularly over the whole image (without any reference to objects and background).

In the semi-automatic mode, the user has to click on some points in the background (hence, the plugin is limited to objects sitting on a background).

Then, in both modes, the background is modelled as a polynomial (the degrees of the polynomial along X and Y are chosen by the user) and the image is divided by the estimated background. The plugin works with 8, 16 or 32 bit-gray level images or color images. For color images, the intensity is corrected. [MORE ...]

Simple global contrast enhancement for gray level images. The gray values are mapped on the interval [mean-3*std, mean+3*std].

In version 2, this plugin can be applied to a stack: all images composing the stack are processed independently.

Image gradient

Gx=[I(x+1,y)-I(x-1,y)]/2 Gy=[I(x,y+1)-I(x,y-1)]/2

One of the several regularized gradient filters for step edge detection (see Chen and Castan, CVGIP, 1992, 54 (2) 112-133)

Being a recursive filter, it runs very fast and independently of the smoothing kernel size. [MORE ...]

Simple segmentation (interactive) [MORE ...]

Multivariate Statistical Analysis (MSA): Principal Component Analysis (PCA), Correspondence Analysis (CA)

Dimensionality reduction: a set of images (either the different components of a multi-component image or a set of related images as in a time series for instance) is compressed in a lower number of images, concentrating most of the information.

[MORE ...]

Performs image segmentation, based on gray-levels histogram, but without having to define gray-level thresholds.

See (Bonnet N., Cutrona J. and Herbin M. Pattern Recognition 2002, 35, 2319-2322) for more details on the procedure. [MORE ...]

Segmentation_of a 2-component_image

Performs 2-component image segmentation (the 2 components may be truly registered components, such as green and red of a color image, or principal components obtained after dimensionality reduction, for instance).

First, the 2D scatterplot is built from the 2 components.

Then, the choice is offered to the user to perform manual (interactive) segmentation (also called Interactive Correlation Partitioning) or automatic segmentation (Automatic Correlation Partitioning).

For Interactive Correlation Partitioning, the user fixes the number (N) of classes he is interested in and then draws N regions of interest (ROI) in the scatterplot. Then, the segmentation of the original image is performed by automatic back-mapping.

For Automatic Correlation Partitioning, the user tries several estimations of the probability density function (pdf) via the Parzen (or kernel) approach. The number of modes of the pdf is assumed to be the number of classes. The user then chooses the kernel parameter (standard deviation of a Gaussian kernel, for instance) which provides the number of classes he is interested in. Then, the parametric space (i.e. the 2D scatterplot) is partitionned according to the watersheds approach, originating from Mathematical Morphology. Finally, the back-mapping of the labels found in the scatterplot allows to segment the original image space.

Version 2: In any case, different coefficients of colocalisation are computed.

[MORE ...]

Reference: Bonnet N. Advances in Electronics and Electron Physics (2000) 114, 1.

Segmentation of a multi-component image

Performs N-component image segmentation (the N components may be truly registered components, such as red, green and blue components of a color image, or principal components obtained after dimensionality reduction, for instance).

At this moment, N is limited to 4, and the computation time for N=4 is prohibitively long. Even for N=3, it is better to work with small scatterplots (32x32x32 for instance).

First, the N-Dimensional scatterplot is built from the N components.

Only Automatic Correlation Partitioning is available (For Interactive Correlation Partitioning with N=2, use the plug-in above).

For Automatic Correlation Partitioning, the user tries several estimations of the probability density function (pdf) via the Parzen (or kernel) approach. The number of modes of the pdf is assumed to be the number of classes. The user then chooses the kernel parameter (standard deviation of a Gaussian kernel, for instance) which provides the number of classes he is interested in. Then, the parametric space (i.e. the ND scatterplot) is partitionned according to the N-dimensional watersheds approach, originating from Mathematical Morphology. Finally, the back-mapping of the labels found in the scatterplot allows to segment the original image space.

[MORE ...]

References: Bonnet N. Advances in Electronics and Electron Physics (2000) 114, 1.

An improved version of this algorithm, using fuzzy logic and probabilistic relaxation concepts, has been developed. See:

* Bonnet N. and Cutrona J.

* Cutrona J., Bonnet N., Herbin M. and Hofer F. Ultramicroscopy (2005) 103, 141.

We hope to translate it for ImageJ (from C++) in the near future.

However, seeds are not automatically obtained. Seeds for objects and background regions are selected interactively by the user.

[MORE ...]

The functions already available in the ImageJ menus are made available through a new toolbar, for a better confort of the user.

[MORE ...]

Developed by Philippe VAUTROT, Maxime PINCHON, Noël BONNET

to the plugins folder, unzip and restart ImageJ.

Developed by Maxime PINCHON, Laetitia PASQUET,
Noël BONNET

to the plugins folder, unzip and restart ImageJ.

Developed by Maxime PINCHON and Noël BONNET

Developed by Maxime PINCHON

to the plugins folder, unzip and restart ImageJ.

Developed by Maxime PINCHON

to the plugins folder, unzip and restart ImageJ

Developed by Maxime PINCHON

**DOWNLOAD
Segmentation_manual_514.zip**

to the plugins folder, unzip and restart ImageJ

Multivariate Statistical Analysis (MSA): Principal Component Analysis (PCA), Correspondence Analysis (CA)

DOWLOAD MSA_514.zipto the plugins folder, unzip and restart ImageJ

Multivariate Statistical Analysis (MSA): Principal Component Analysis (PCA), Correspondence Analysis (CA)

Developed
by Gael LALIRE, Benjamin PROUVOST and Noel BONNET

to the plugins folder, unzip and restart ImageJ

Developed by Maxime PINCHON and Noël BONNET

Version 2 by Cedric GILLET and Noël BONNET

to the plugins folder, unzip and restart ImageJ.

Segmentation_2_component_image

**Watersheds-based image
segmentation**

Developed by Laetitia PASQUET and
Noël BONNET

DOWNLOAD
Segmentation_2_component_image_514.zip

to the plugins folder, unzip and restart ImageJ

Segmentation_max4_component_image

to the plugins folder, unzip and restart ImageJ

to the plugins folder, unzip and restart ImageJ

Segmentation_max4_component_image

Developed by Laetitia PASQUET and
Noël BONNET

DOWNLOAD
Segmentation_max4_component_image_514.zipto the plugins folder, unzip and restart ImageJ

Developed by Maxime PINCHON and Noël BONNET

to the plugins folder, unzip and restart ImageJ.

Developed by Maxime PINCHON

Noël Bonnet July 2004, October 2004, June 2005, January 2006, July 2006, April 2007