zeyut/food-segmentation

This is a project that segmented foods in photographs via an active contour algorithm

Food-Segmentation

Introduction

This is a project that segmented foods in photographs via an active contour algorithm. The project is the semester project of Professor Hoover's course ECE6310 in Clemson University.

Implementation

The goal is to implement an active contour algorithm that allows a user to semi-automatically segment food items in a photograph of multiple foods on a dinner plate. The program must load a color PNM image but only needs to display the greyscale version (average of the three color bands). Users have the following options: left-click to draw around a food item that subsequently automatically shrinks to wrap to the food boundary; right-click a point within a food item that subsequently automatically grows a contour to its outer boundary; shift-click (either button) to manually drag a contour point to a new location.

There are several energy terms being considered.Three internal energy terms includes the distance between points, the deviation from the average distance between points, and the distance to the centroid of the current contour. Three external energy terms includes the image Sobel edge gradient magnitude, the average intensity within the contour, and the Canny edge gradient magnitude.

Besides, the project files integrates a region grow algorithm that could grow a region from a point in a photograph. However, the region grow algorithm only works well on objects with obvious edges.

Dependencies

The project requires:

• Window 8 SDK
• Visual Studio 2017 or above

Files

• Emerge.sln: the structure file for VS.
• Emerge.exe: the executable file of the project.
• main.c: the C code of the GUI.
• ActiveContour.c: the C code of the active contour algorithm
• RegionGrow.c: the C code of the region grow algorithm.

Examples

Here are some examples of food segmentation in growing mode.

Here are some examples of food segmentation in shrinking mode.