How to Best Analyze an Image of Ice Floes with Object Differentiation and Distribution? (Switching from ImageJ to MatLab)

Hello,

I would like to refine a process that I am currently using to analyze photos of a channel of broken ice pieces. I have used MatLab before but I am totally new to the image processing and analysis capabilities of the program.

My goal is to take a photo like this below:

... And produce a histogram of floe size distribution that includes total area of ice.

Currently I am using ImageJ to do this and the process goes like this:

• Downsize the photo
• Change to black and white binary
• Change to 8-bit type
• Set pixel to real length ratio
• Use watershed function to determine the boundaries of overlapping floe pieces

I have tried to directly translate the process into matlab but I keep getting the feeling that I am not doing it optimally. Any direction and help would be greatly appreciated, including information about how to approach the problem down to specific commands that are useful for this task.

[Matt Kindig]

Hi Andrew,

Honestly, the best way to get familiar with Matlab's Image Processing Toolbox capabilities are the demos for it. See Contents -> Image Processing Toolbox -> Demos in the Help browser. There are a number for image segmentation, including one that specifically demonstrates using the watershed function.

[Andrew]

Thanks, I'll start having a look at those now.

Answer 1

[Image Analyst]

Did it even work in ImageJ? What does your code look like?

[Andrew]

It did, but I didn't code it into imageJ, just used the dropdown menus as I went along. Its a leftover procedure from a previous employee which is why I want to clean it up.

Although ImageJ did not give a histogram, it gave a mess of data that had to be put into excel.