Blob Batch Button

These commands are for collecting measurements of particles in a folder of images. This is done in two or three stages: 1) selecting thresholds for the images, 2) collecting the statistics for the images, and 3) making a summary spreadsheet file, if desired, for the statistics for each image. This last file is used by the Fields in Circle Tool, to colorize a mosaic of fields by number of particles, greatest particle, etc. The thresholds are saved in a tab delimited text file, one image per line (or row). The blob statistics (particle measurements) are saved in a similar text file, one blob per line. The summary statistics are saved in a text file, one image per line. These text files are saved in a folder named <image folder name>-lx, which follows the folder of images in a directory listing, for example, if the images are in a folder called "Fields", then the directory structure looks like this:

The threshold files and statistics files are named by the thresholding method, such as "fixed-thresholds.txt" and "fixed-stats.txt" for the one-threhold-fits-all method, and "manual-thresholds.txt" and "manual-stats.txt" for the manually-chosen-threshold-for-each-image method. These text files can be edited with a text editor, and imported into a spreadsheet. The blob stats and summary files are easier to read when imported into Excel. The order of the images listed in the threshold file is not important.

Make Threshold File -

Manual: Displays images one by one. You choose the thresholds.
- As each image is displayed, it is zoomed and scrolled to the same place as the previous image, and the threshold slider is initialized to the position and thresholds that you recorded for the previous image.
- If partially completed and restarted, this routine skips images already listed in the threshold file. If you wish to edit a threshold, you can delete the appropriate line in the file before running the "Manual Threshold" command.
Fixed: Displays the image on which you click to select the folder of images. The thresholds you select for this image are applied to all.
Auto-1: Displays the image on which you click to select the folder of images. You select the threshold for this image. Analyzes the background of each image using a routine that counts the isolated thresholded pixels for particular threshold settings. Attempts to select a threshold for each image to give a number of single pixel blobs that is equal to or less than the number for the image for which you selected the threshold. Assumes background is relatively smooth and dark, and covers all but a few percent of each image. Also assumes than no particles of interest are one pixel big, at this magnification.

Sample output:

Image T1 T2 1PCounts Class
Field001 30 78 6534 normal
Field002 31 84 5139 normal
Field003 31 48 5611 normal

Image - image file name less extension.

T1 - Threshold found by auto1 method. All pixels with this value or brighter are candidates for single pixel blobs. T2 is the maximum pixel intensity for the image, and is included here just for your information.

1PCounts - number of single pixel blobs for threshold T1. All of these pixels are isolated, and have pixel values between T1 and T2 inclusive.

Class - either normal or dip. The method starts with a threshold that is too low (too many single pixel blobs) and increases the threshold until the number is equal to or less than the number in the image for which you selected the threshold. Dip means that the number of single pixel blobs has increased for this new threshold, rather than decreasing as expected. This may be caused by large particles or large areas of brighter background -- the image is probably anomalous.

Auto2 Displays the image on which you click to select the folder of images. You select the threshold for this image. Does a quadratic fit to the histogram for each image, and assumes that the proper threshold will be C + xW, where C is the center of the histogram peak, W is the right half width and x is a factor calculated from your threshold selection of the above image. Assumes that differences between images are due to small changes in offset and gain. Assumes that the major peak in the histogram is due to background pixels, and that the texture of the background is not clipped and thus represented by only one or two gray levels.
Auto2a Same as Auto2, except that it uses the center of the backtround peak of the histogram plus a constant offset. Auto2 uses an offset proportional to the width of the background peak.
Hist2 Makes a threshold-like file with the histogram background peak center and width. These values are not thresholds per-se, but are used in the Auto2 methods. The file is useful for colorizing the Fields In Circle Tool after loading a series of images from an Aspex* run.
Auto3 Requires no user input such as choosing a threshold from a sample image. Sets the threshold at an estimate of the upper brightness limit for the background. In short, the method assumes:
- Most of the image is background, in other words the tallest peak of the image represents background.
- Objects are brighter than the background.
- Background is realtively smooth. Even dim and small blobs that are just slightly brighter than the background are candidates to be objects.
The steps to find the threshold are:
- Calculate the gray level histogram of the image.
- Find the tallest peak in the histogram.
- Fit a parabola to the top half of the peak.
- Find the right zero intercept (x-axis) for the parabola.
- A skirt of the background peak extends to the right of the zero intercept. From the first point to the right of the zero intercept, which is the highest point on the skirt, fit a straight line to the top half of the skirt, in other words to all points on the skirt from the first point to the point just before the value of the point is less than half that of the first point.
- The threshold is the zero intercept (x-axis) of this line.

Uses the images you select and a threshold file you have made.
Preprocessing
- (new 2/13/09) Now has option to use a process-images.txt file to preprocess each image before blobbing. The file goes in the *-lx folder. For cropping the images, it need have only one line, for example "crop 0 0 820 820". The four numbers following "crop" are the bounding rectangle for blobbing the image. You can get these numbers by drawing the bounding rectangle with the Select menu, and cut & pasting the numbers from either of these displays:
  - Select / Show Rectangle Info / Print
  - Select / Set by Typing / X1 Y1 X2 Y2
- Other preprocessing commands are available, which are also used with the Process Folder Tool.
Can take 10 - 20 minutes for a run with over 300 images, 1 - 100 particles per image.
Makes one long text file, spreadsheet compatable, with each blob in one row: Image Name, Blob Index, stat1, stat2, ...
The file is named "manual-stats.txt", or "fixed-stats.txt", etc., as appropriate.
Use the Stats Button, just to the left of the Blob Button, to choose which blob measurements to record.

Blob Batch Button

Batch Commands